# Standard libraries
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import time

# Scikit-learn
from sklearn.metrics import classification_report

# TensorFlow / Keras
import tensorflow as tf

# Project modules
import sys
sys.path.insert(0, ".")
from src.config import (
    RANDOM_SEED, SPLIT_RANDOM_STATE, TEST_SIZE,
    THRESHOLD, EPOCHS, BATCH_SIZE, TRAIN_PATH, TEST_PATH,
)
from src.preprocessing import (
    load_data, train_val_split, fit_transform_imputer, compute_class_weights,
)
from src.models import build_model
from src.evaluation import (
    model_performance_classification, compare_models,
    plot_history, print_classification_report,
)

# Suppress warnings
import warnings
warnings.filterwarnings("ignore")

# Reproducibility seeds
np.random.seed(RANDOM_SEED)
tf.random.set_seed(RANDOM_SEED)

# Display settings
pd.set_option("display.max_columns", 50)
sns.set_style("whitegrid")

print(f"NumPy:      {np.__version__}")
print(f"Pandas:     {pd.__version__}")
print(f"TensorFlow: {tf.__version__}")
print(f"Random Seed: {RANDOM_SEED}")
print(f"Threshold:   {THRESHOLD}")

NumPy:      2.0.2
Pandas:     3.0.0
TensorFlow: 2.18.0
Random Seed: 42
Threshold:   0.5

# Load data
data, data_test = load_data(TRAIN_PATH, TEST_PATH)

print(f"Training data shape: {data.shape}")
print(f"Test data shape:     {data_test.shape}")

Training data shape: (20000, 41)
Test data shape:     (5000, 41)

# First 5 rows of training data
data.head()

# First 5 rows of test data
data_test.head()

# Data types
data.dtypes

V1        float64
V2        float64
V3        float64
V4        float64
V5        float64
V6        float64
V7        float64
V8        float64
V9        float64
V10       float64
V11       float64
V12       float64
V13       float64
V14       float64
V15       float64
V16       float64
V17       float64
V18       float64
V19       float64
V20       float64
V21       float64
V22       float64
V23       float64
V24       float64
V25       float64
V26       float64
V27       float64
V28       float64
V29       float64
V30       float64
V31       float64
V32       float64
V33       float64
V34       float64
V35       float64
V36       float64
V37       float64
V38       float64
V39       float64
V40       float64
Target    float64
dtype: object

# Check for duplicates
data.duplicated().sum()

np.int64(0)

# Missing values — training data
data.isnull().sum()

V1        18
V2        18
V3         0
V4         0
V5         0
V6         0
V7         0
V8         0
V9         0
V10        0
V11        0
V12        0
V13        0
V14        0
V15        0
V16        0
V17        0
V18        0
V19        0
V20        0
V21        0
V22        0
V23        0
V24        0
V25        0
V26        0
V27        0
V28        0
V29        0
V30        0
V31        0
V32        0
V33        0
V34        0
V35        0
V36        0
V37        0
V38        0
V39        0
V40        0
Target     0
dtype: int64

# Missing values — test data
data_test.isnull().sum()

V1        5
V2        6
V3        0
V4        0
V5        0
V6        0
V7        0
V8        0
V9        0
V10       0
V11       0
V12       0
V13       0
V14       0
V15       0
V16       0
V17       0
V18       0
V19       0
V20       0
V21       0
V22       0
V23       0
V24       0
V25       0
V26       0
V27       0
V28       0
V29       0
V30       0
V31       0
V32       0
V33       0
V34       0
V35       0
V36       0
V37       0
V38       0
V39       0
V40       0
Target    0
dtype: int64

# Statistical summary
data.describe()

def histogram_boxplot(data, feature, figsize=(12, 7), kde=False, bins=None):
    """Boxplot and histogram combined."""
    f2, (ax_box2, ax_hist2) = plt.subplots(
        nrows=2,
        sharex=True,
        gridspec_kw={"height_ratios": (0.25, 0.75)},
        figsize=figsize,
    )
    sns.boxplot(data=data, x=feature, ax=ax_box2, showmeans=True, color="violet")
    if bins:
        sns.histplot(data=data, x=feature, kde=kde, ax=ax_hist2, bins=bins)
    else:
        sns.histplot(data=data, x=feature, kde=kde, ax=ax_hist2)
    ax_hist2.axvline(data[feature].mean(), color="green", linestyle="--")
    ax_hist2.axvline(data[feature].median(), color="black", linestyle="-")
    plt.show()

# Distributions for all features
for feature in data.columns:
    histogram_boxplot(data, feature, figsize=(12, 7), kde=False, bins=None)

# Target distribution — training data
print("Training set target proportions:")
print(data["Target"].value_counts(normalize=True))
print()
print("Test set target proportions:")
print(data_test["Target"].value_counts(normalize=True))

Training set target proportions:
Target
0.0    0.9445
1.0    0.0555
Name: proportion, dtype: float64

Test set target proportions:
Target
0.0    0.9436
1.0    0.0564
Name: proportion, dtype: float64

cols_list = data.select_dtypes(include=np.number).columns.tolist()
cols_list.remove("Target")

plt.figure(figsize=(20, 20))
sns.heatmap(
    data[cols_list].corr(), annot=True, vmin=-1, vmax=1, fmt=".2f", cmap="Spectral"
)
plt.show()

# Top separating features — rank by absolute difference in class means
class_means = data.groupby("Target").mean(numeric_only=True)
mean_diff = (class_means.loc[1] - class_means.loc[0]).abs().sort_values(ascending=False)
top_features = mean_diff.head(6).index.tolist()

fig, axes = plt.subplots(2, 3, figsize=(16, 8))
for ax, feat in zip(axes.ravel(), top_features):
    for label, color in [(0, "steelblue"), (1, "tomato")]:
        subset = data[data["Target"] == label][feat].dropna()
        ax.hist(subset, bins=40, alpha=0.5, label=f"Class {label}", color=color, density=True)
    ax.set_title(feat)
    ax.legend()
fig.suptitle("Top 6 Features by Class-Mean Separation", fontsize=14)
plt.tight_layout()
plt.show()

# Separate features and target
X = data.drop(columns=["Target"])
y = data["Target"]

# Train / Validation split (stratified)
X_train, X_val, y_train, y_val = train_val_split(
    X, y, test_size=TEST_SIZE, random_state=SPLIT_RANDOM_STATE
)

print(f"X_train shape: {X_train.shape}")
print(f"X_val shape:   {X_val.shape}")

X_train shape: (16000, 40)
X_val shape:   (4000, 40)

# Separate test features and target
X_test = data_test.drop(columns=["Target"], errors="ignore")
y_test = data_test["Target"] if "Target" in data_test.columns else None
print(f"X_test shape:  {X_test.shape}")

X_test shape:  (5000, 40)

# Impute missing values (median, fit on train only)
X_train, X_val, X_test, imputer = fit_transform_imputer(X_train, X_val, X_test)

# Verify no missing values remain
print(X_train.isna().sum().sum(), X_val.isna().sum().sum(), X_test.isna().sum().sum())

0 0 0

# Convert targets to numpy arrays
y_train = y_train.to_numpy()
y_val = y_val.to_numpy()
y_test = y_test.to_numpy()

print(f"Epochs:     {EPOCHS}")
print(f"Batch size: {BATCH_SIZE}")
print(f"Threshold:  {THRESHOLD}")
print(f"Input dim:  {X_train.shape[1]}")

Epochs:     50
Batch size: 32
Threshold:  0.5
Input dim:  40

# Compute class weights (used by models 3 and 6)
cw_dict = compute_class_weights(y_train)
print(f"Class weights: {cw_dict}")

Class weights: {0: np.float64(1.0587612493382743), 1: np.float64(18.01801801801802)}

# Storage for performance results
perf_train = {}
perf_val = {}

config_0 = {"layers": [7], "activation": "relu"}
model_0 = build_model(X_train.shape[1], config_0, "sgd")
model_0.summary()

Model: "sequential"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ dense (Dense)                   │ (None, 7)              │           287 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 1)              │             8 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 295 (1.15 KB)

 Trainable params: 295 (1.15 KB)

 Non-trainable params: 0 (0.00 B)

start = time.time()
history_0 = model_0.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    batch_size=BATCH_SIZE,
    epochs=EPOCHS,
)
print(f"Time taken: {time.time() - start:.1f}s")

Epoch 1/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 1s 617us/step - accuracy: 0.9534 - loss: 0.1701 - val_accuracy: 0.9668 - val_loss: 0.1256
Epoch 2/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 487us/step - accuracy: 0.9674 - loss: 0.1173 - val_accuracy: 0.9710 - val_loss: 0.1137
Epoch 3/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 479us/step - accuracy: 0.9701 - loss: 0.1069 - val_accuracy: 0.9728 - val_loss: 0.1066
Epoch 4/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 479us/step - accuracy: 0.9721 - loss: 0.1006 - val_accuracy: 0.9743 - val_loss: 0.1022
Epoch 5/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 482us/step - accuracy: 0.9737 - loss: 0.0961 - val_accuracy: 0.9737 - val_loss: 0.0986
Epoch 6/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 484us/step - accuracy: 0.9751 - loss: 0.0924 - val_accuracy: 0.9747 - val_loss: 0.0959
Epoch 7/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 480us/step - accuracy: 0.9756 - loss: 0.0891 - val_accuracy: 0.9758 - val_loss: 0.0935
Epoch 8/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 484us/step - accuracy: 0.9768 - loss: 0.0861 - val_accuracy: 0.9780 - val_loss: 0.0911
Epoch 9/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 478us/step - accuracy: 0.9781 - loss: 0.0832 - val_accuracy: 0.9787 - val_loss: 0.0889
Epoch 10/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 477us/step - accuracy: 0.9793 - loss: 0.0802 - val_accuracy: 0.9800 - val_loss: 0.0869
Epoch 11/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 482us/step - accuracy: 0.9803 - loss: 0.0774 - val_accuracy: 0.9818 - val_loss: 0.0849
Epoch 12/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 478us/step - accuracy: 0.9814 - loss: 0.0750 - val_accuracy: 0.9830 - val_loss: 0.0831
Epoch 13/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 479us/step - accuracy: 0.9824 - loss: 0.0729 - val_accuracy: 0.9837 - val_loss: 0.0813
Epoch 14/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 474us/step - accuracy: 0.9833 - loss: 0.0711 - val_accuracy: 0.9840 - val_loss: 0.0798
Epoch 15/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 477us/step - accuracy: 0.9841 - loss: 0.0697 - val_accuracy: 0.9843 - val_loss: 0.0784
Epoch 16/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 475us/step - accuracy: 0.9845 - loss: 0.0684 - val_accuracy: 0.9845 - val_loss: 0.0771
Epoch 17/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 670us/step - accuracy: 0.9851 - loss: 0.0671 - val_accuracy: 0.9852 - val_loss: 0.0760
Epoch 18/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 476us/step - accuracy: 0.9855 - loss: 0.0661 - val_accuracy: 0.9855 - val_loss: 0.0749
Epoch 19/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 477us/step - accuracy: 0.9859 - loss: 0.0651 - val_accuracy: 0.9858 - val_loss: 0.0740
Epoch 20/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 477us/step - accuracy: 0.9864 - loss: 0.0642 - val_accuracy: 0.9860 - val_loss: 0.0731
Epoch 21/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 476us/step - accuracy: 0.9867 - loss: 0.0633 - val_accuracy: 0.9862 - val_loss: 0.0722
Epoch 22/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 478us/step - accuracy: 0.9869 - loss: 0.0625 - val_accuracy: 0.9868 - val_loss: 0.0714
Epoch 23/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 473us/step - accuracy: 0.9869 - loss: 0.0617 - val_accuracy: 0.9872 - val_loss: 0.0707
Epoch 24/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 478us/step - accuracy: 0.9868 - loss: 0.0610 - val_accuracy: 0.9875 - val_loss: 0.0701
Epoch 25/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 483us/step - accuracy: 0.9869 - loss: 0.0603 - val_accuracy: 0.9877 - val_loss: 0.0695
Epoch 26/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 479us/step - accuracy: 0.9869 - loss: 0.0598 - val_accuracy: 0.9877 - val_loss: 0.0690
Epoch 27/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 477us/step - accuracy: 0.9872 - loss: 0.0593 - val_accuracy: 0.9877 - val_loss: 0.0687
Epoch 28/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 480us/step - accuracy: 0.9874 - loss: 0.0589 - val_accuracy: 0.9880 - val_loss: 0.0684
Epoch 29/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 474us/step - accuracy: 0.9876 - loss: 0.0585 - val_accuracy: 0.9880 - val_loss: 0.0681
Epoch 30/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 473us/step - accuracy: 0.9877 - loss: 0.0582 - val_accuracy: 0.9880 - val_loss: 0.0678
Epoch 31/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 475us/step - accuracy: 0.9878 - loss: 0.0578 - val_accuracy: 0.9880 - val_loss: 0.0676
Epoch 32/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 477us/step - accuracy: 0.9881 - loss: 0.0576 - val_accuracy: 0.9883 - val_loss: 0.0674
Epoch 33/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 496us/step - accuracy: 0.9881 - loss: 0.0573 - val_accuracy: 0.9885 - val_loss: 0.0673
Epoch 34/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 476us/step - accuracy: 0.9883 - loss: 0.0570 - val_accuracy: 0.9885 - val_loss: 0.0672
Epoch 35/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 473us/step - accuracy: 0.9884 - loss: 0.0567 - val_accuracy: 0.9885 - val_loss: 0.0671
Epoch 36/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 473us/step - accuracy: 0.9883 - loss: 0.0565 - val_accuracy: 0.9885 - val_loss: 0.0670
Epoch 37/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 474us/step - accuracy: 0.9884 - loss: 0.0563 - val_accuracy: 0.9885 - val_loss: 0.0670
Epoch 38/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 474us/step - accuracy: 0.9884 - loss: 0.0561 - val_accuracy: 0.9885 - val_loss: 0.0669
Epoch 39/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 472us/step - accuracy: 0.9884 - loss: 0.0560 - val_accuracy: 0.9885 - val_loss: 0.0669
Epoch 40/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 477us/step - accuracy: 0.9884 - loss: 0.0558 - val_accuracy: 0.9883 - val_loss: 0.0668
Epoch 41/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 476us/step - accuracy: 0.9883 - loss: 0.0556 - val_accuracy: 0.9883 - val_loss: 0.0668
Epoch 42/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 475us/step - accuracy: 0.9883 - loss: 0.0555 - val_accuracy: 0.9883 - val_loss: 0.0667
Epoch 43/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 478us/step - accuracy: 0.9884 - loss: 0.0553 - val_accuracy: 0.9883 - val_loss: 0.0667
Epoch 44/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 476us/step - accuracy: 0.9884 - loss: 0.0552 - val_accuracy: 0.9883 - val_loss: 0.0667
Epoch 45/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 475us/step - accuracy: 0.9884 - loss: 0.0551 - val_accuracy: 0.9883 - val_loss: 0.0667
Epoch 46/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 630us/step - accuracy: 0.9885 - loss: 0.0550 - val_accuracy: 0.9877 - val_loss: 0.0667
Epoch 47/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 476us/step - accuracy: 0.9886 - loss: 0.0548 - val_accuracy: 0.9880 - val_loss: 0.0667
Epoch 48/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 475us/step - accuracy: 0.9886 - loss: 0.0548 - val_accuracy: 0.9877 - val_loss: 0.0667
Epoch 49/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 477us/step - accuracy: 0.9885 - loss: 0.0547 - val_accuracy: 0.9877 - val_loss: 0.0667
Epoch 50/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 475us/step - accuracy: 0.9886 - loss: 0.0546 - val_accuracy: 0.9880 - val_loss: 0.0667
Time taken: 13.0s

plot_history(history_0, "loss")

perf_train["Model 0"] = model_performance_classification(model_0, X_train, y_train)
perf_val["Model 0"] = model_performance_classification(model_0, X_val, y_val)
print("Train:"); display(perf_train["Model 0"])
print("Validation:"); display(perf_val["Model 0"])

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 227us/step
125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 275us/step
Train:

Validation:

print_classification_report(model_0, X_train, y_train, "Train", "Model_0")
print_classification_report(model_0, X_val, y_val, "Validation", "Model_0")

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 227us/step
Classification Report - Train data Model_0

              precision    recall  f1-score   support

         0.0       0.99      1.00      0.99     15112
         1.0       0.97      0.82      0.89       888

    accuracy                           0.99     16000
   macro avg       0.98      0.91      0.94     16000
weighted avg       0.99      0.99      0.99     16000

125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 269us/step
Classification Report - Validation data Model_0

              precision    recall  f1-score   support

         0.0       0.99      1.00      0.99      3778
         1.0       0.97      0.81      0.88       222

    accuracy                           0.99      4000
   macro avg       0.98      0.90      0.94      4000
weighted avg       0.99      0.99      0.99      4000

config_1 = {"layers": [14, 7], "activation": "relu"}
model_1 = build_model(X_train.shape[1], config_1, "sgd")
model_1.summary()

Model: "sequential"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ dense (Dense)                   │ (None, 14)             │           574 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 7)              │           105 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_2 (Dense)                 │ (None, 1)              │             8 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 687 (2.68 KB)

 Trainable params: 687 (2.68 KB)

 Non-trainable params: 0 (0.00 B)

start = time.time()
history_1 = model_1.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    batch_size=BATCH_SIZE,
    epochs=EPOCHS,
)
print(f"Time taken: {time.time() - start:.1f}s")

Epoch 1/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 639us/step - accuracy: 0.9467 - loss: 0.1651 - val_accuracy: 0.9607 - val_loss: 0.1233
Epoch 2/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 498us/step - accuracy: 0.9648 - loss: 0.1183 - val_accuracy: 0.9700 - val_loss: 0.1082
Epoch 3/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 495us/step - accuracy: 0.9715 - loss: 0.1038 - val_accuracy: 0.9758 - val_loss: 0.0982
Epoch 4/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 496us/step - accuracy: 0.9771 - loss: 0.0935 - val_accuracy: 0.9805 - val_loss: 0.0910
Epoch 5/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 492us/step - accuracy: 0.9796 - loss: 0.0858 - val_accuracy: 0.9833 - val_loss: 0.0856
Epoch 6/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 491us/step - accuracy: 0.9816 - loss: 0.0798 - val_accuracy: 0.9840 - val_loss: 0.0811
Epoch 7/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 487us/step - accuracy: 0.9824 - loss: 0.0750 - val_accuracy: 0.9858 - val_loss: 0.0777
Epoch 8/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 488us/step - accuracy: 0.9834 - loss: 0.0713 - val_accuracy: 0.9865 - val_loss: 0.0751
Epoch 9/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 495us/step - accuracy: 0.9846 - loss: 0.0682 - val_accuracy: 0.9868 - val_loss: 0.0731
Epoch 10/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 488us/step - accuracy: 0.9855 - loss: 0.0656 - val_accuracy: 0.9865 - val_loss: 0.0718
Epoch 11/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9862 - loss: 0.0637 - val_accuracy: 0.9862 - val_loss: 0.0708
Epoch 12/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 487us/step - accuracy: 0.9867 - loss: 0.0621 - val_accuracy: 0.9865 - val_loss: 0.0698
Epoch 13/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 492us/step - accuracy: 0.9871 - loss: 0.0606 - val_accuracy: 0.9870 - val_loss: 0.0692
Epoch 14/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 492us/step - accuracy: 0.9875 - loss: 0.0594 - val_accuracy: 0.9870 - val_loss: 0.0686
Epoch 15/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 487us/step - accuracy: 0.9874 - loss: 0.0584 - val_accuracy: 0.9870 - val_loss: 0.0683
Epoch 16/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 487us/step - accuracy: 0.9877 - loss: 0.0574 - val_accuracy: 0.9870 - val_loss: 0.0680
Epoch 17/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9883 - loss: 0.0564 - val_accuracy: 0.9870 - val_loss: 0.0678
Epoch 18/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 648us/step - accuracy: 0.9885 - loss: 0.0555 - val_accuracy: 0.9870 - val_loss: 0.0676
Epoch 19/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 491us/step - accuracy: 0.9888 - loss: 0.0548 - val_accuracy: 0.9868 - val_loss: 0.0675
Epoch 20/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 488us/step - accuracy: 0.9890 - loss: 0.0541 - val_accuracy: 0.9868 - val_loss: 0.0675
Epoch 21/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 489us/step - accuracy: 0.9893 - loss: 0.0534 - val_accuracy: 0.9870 - val_loss: 0.0675
Epoch 22/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 520us/step - accuracy: 0.9894 - loss: 0.0528 - val_accuracy: 0.9877 - val_loss: 0.0674
Epoch 23/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 488us/step - accuracy: 0.9893 - loss: 0.0522 - val_accuracy: 0.9877 - val_loss: 0.0674
Epoch 24/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9894 - loss: 0.0517 - val_accuracy: 0.9877 - val_loss: 0.0675
Epoch 25/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 491us/step - accuracy: 0.9896 - loss: 0.0512 - val_accuracy: 0.9877 - val_loss: 0.0674
Epoch 26/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9897 - loss: 0.0508 - val_accuracy: 0.9877 - val_loss: 0.0675
Epoch 27/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 488us/step - accuracy: 0.9896 - loss: 0.0504 - val_accuracy: 0.9875 - val_loss: 0.0675
Epoch 28/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 492us/step - accuracy: 0.9899 - loss: 0.0501 - val_accuracy: 0.9877 - val_loss: 0.0675
Epoch 29/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 485us/step - accuracy: 0.9899 - loss: 0.0497 - val_accuracy: 0.9877 - val_loss: 0.0675
Epoch 30/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 495us/step - accuracy: 0.9898 - loss: 0.0494 - val_accuracy: 0.9880 - val_loss: 0.0675
Epoch 31/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 492us/step - accuracy: 0.9898 - loss: 0.0491 - val_accuracy: 0.9880 - val_loss: 0.0676
Epoch 32/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9899 - loss: 0.0487 - val_accuracy: 0.9880 - val_loss: 0.0676
Epoch 33/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9899 - loss: 0.0484 - val_accuracy: 0.9880 - val_loss: 0.0676
Epoch 34/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 583us/step - accuracy: 0.9902 - loss: 0.0481 - val_accuracy: 0.9883 - val_loss: 0.0677
Epoch 35/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 487us/step - accuracy: 0.9902 - loss: 0.0478 - val_accuracy: 0.9885 - val_loss: 0.0677
Epoch 36/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9902 - loss: 0.0475 - val_accuracy: 0.9883 - val_loss: 0.0677
Epoch 37/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 493us/step - accuracy: 0.9901 - loss: 0.0472 - val_accuracy: 0.9883 - val_loss: 0.0677
Epoch 38/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9902 - loss: 0.0470 - val_accuracy: 0.9883 - val_loss: 0.0678
Epoch 39/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 488us/step - accuracy: 0.9904 - loss: 0.0467 - val_accuracy: 0.9880 - val_loss: 0.0678
Epoch 40/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 489us/step - accuracy: 0.9903 - loss: 0.0465 - val_accuracy: 0.9885 - val_loss: 0.0678
Epoch 41/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 487us/step - accuracy: 0.9902 - loss: 0.0463 - val_accuracy: 0.9883 - val_loss: 0.0678
Epoch 42/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 489us/step - accuracy: 0.9906 - loss: 0.0460 - val_accuracy: 0.9885 - val_loss: 0.0677
Epoch 43/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 486us/step - accuracy: 0.9906 - loss: 0.0459 - val_accuracy: 0.9885 - val_loss: 0.0676
Epoch 44/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 489us/step - accuracy: 0.9906 - loss: 0.0457 - val_accuracy: 0.9885 - val_loss: 0.0677
Epoch 45/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 487us/step - accuracy: 0.9906 - loss: 0.0455 - val_accuracy: 0.9885 - val_loss: 0.0676
Epoch 46/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 489us/step - accuracy: 0.9906 - loss: 0.0453 - val_accuracy: 0.9883 - val_loss: 0.0675
Epoch 47/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 491us/step - accuracy: 0.9905 - loss: 0.0452 - val_accuracy: 0.9885 - val_loss: 0.0676
Epoch 48/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 658us/step - accuracy: 0.9906 - loss: 0.0450 - val_accuracy: 0.9885 - val_loss: 0.0675
Epoch 49/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9908 - loss: 0.0449 - val_accuracy: 0.9885 - val_loss: 0.0675
Epoch 50/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 490us/step - accuracy: 0.9909 - loss: 0.0447 - val_accuracy: 0.9883 - val_loss: 0.0675
Time taken: 13.3s

plot_history(history_1, "loss")

perf_train["Model 1"] = model_performance_classification(model_1, X_train, y_train)
perf_val["Model 1"] = model_performance_classification(model_1, X_val, y_val)
print("Train:"); display(perf_train["Model 1"])
print("Validation:"); display(perf_val["Model 1"])

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 330us/step
125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 299us/step
Train:

Validation:

print_classification_report(model_1, X_train, y_train, "Train", "Model_1")
print_classification_report(model_1, X_val, y_val, "Validation", "Model_1")

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 232us/step
Classification Report - Train data Model_1

              precision    recall  f1-score   support

         0.0       0.99      1.00      1.00     15112
         1.0       0.98      0.86      0.92       888

    accuracy                           0.99     16000
   macro avg       0.99      0.93      0.96     16000
weighted avg       0.99      0.99      0.99     16000

125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 274us/step
Classification Report - Validation data Model_1

              precision    recall  f1-score   support

         0.0       0.99      1.00      0.99      3778
         1.0       0.96      0.82      0.89       222

    accuracy                           0.99      4000
   macro avg       0.98      0.91      0.94      4000
weighted avg       0.99      0.99      0.99      4000

config_2 = {"layers": [32, 16, 8], "activation": "relu", "dropout_rate": 0.5}
model_2 = build_model(X_train.shape[1], config_2, "sgd")
model_2.summary()

Model: "sequential"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ dense (Dense)                   │ (None, 32)             │         1,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dropout (Dropout)               │ (None, 32)             │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 16)             │           528 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_2 (Dense)                 │ (None, 8)              │           136 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_3 (Dense)                 │ (None, 1)              │             9 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 1,985 (7.75 KB)

 Trainable params: 1,985 (7.75 KB)

 Non-trainable params: 0 (0.00 B)

start = time.time()
history_2 = model_2.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    batch_size=BATCH_SIZE,
    epochs=EPOCHS,
)
print(f"Time taken: {time.time() - start:.1f}s")

Epoch 1/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 1s 659us/step - accuracy: 0.9433 - loss: 0.2214 - val_accuracy: 0.9578 - val_loss: 0.1380
Epoch 2/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9502 - loss: 0.1678 - val_accuracy: 0.9620 - val_loss: 0.1198
Epoch 3/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 772us/step - accuracy: 0.9542 - loss: 0.1464 - val_accuracy: 0.9670 - val_loss: 0.1107
Epoch 4/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9592 - loss: 0.1376 - val_accuracy: 0.9715 - val_loss: 0.1031
Epoch 5/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 504us/step - accuracy: 0.9605 - loss: 0.1313 - val_accuracy: 0.9712 - val_loss: 0.1002
Epoch 6/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 503us/step - accuracy: 0.9624 - loss: 0.1264 - val_accuracy: 0.9745 - val_loss: 0.0960
Epoch 7/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9645 - loss: 0.1226 - val_accuracy: 0.9743 - val_loss: 0.0947
Epoch 8/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9638 - loss: 0.1202 - val_accuracy: 0.9750 - val_loss: 0.0921
Epoch 9/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 504us/step - accuracy: 0.9664 - loss: 0.1127 - val_accuracy: 0.9755 - val_loss: 0.0893
Epoch 10/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9688 - loss: 0.1085 - val_accuracy: 0.9772 - val_loss: 0.0869
Epoch 11/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 518us/step - accuracy: 0.9706 - loss: 0.1021 - val_accuracy: 0.9790 - val_loss: 0.0845
Epoch 12/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 496us/step - accuracy: 0.9712 - loss: 0.1027 - val_accuracy: 0.9803 - val_loss: 0.0816
Epoch 13/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9719 - loss: 0.1016 - val_accuracy: 0.9818 - val_loss: 0.0803
Epoch 14/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9725 - loss: 0.1001 - val_accuracy: 0.9833 - val_loss: 0.0788
Epoch 15/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9740 - loss: 0.0959 - val_accuracy: 0.9833 - val_loss: 0.0767
Epoch 16/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 500us/step - accuracy: 0.9747 - loss: 0.0953 - val_accuracy: 0.9830 - val_loss: 0.0762
Epoch 17/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 500us/step - accuracy: 0.9744 - loss: 0.0913 - val_accuracy: 0.9827 - val_loss: 0.0751
Epoch 18/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 663us/step - accuracy: 0.9748 - loss: 0.0940 - val_accuracy: 0.9837 - val_loss: 0.0739
Epoch 19/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 496us/step - accuracy: 0.9758 - loss: 0.0887 - val_accuracy: 0.9837 - val_loss: 0.0732
Epoch 20/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9774 - loss: 0.0882 - val_accuracy: 0.9840 - val_loss: 0.0718
Epoch 21/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 498us/step - accuracy: 0.9779 - loss: 0.0848 - val_accuracy: 0.9840 - val_loss: 0.0711
Epoch 22/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9767 - loss: 0.0865 - val_accuracy: 0.9843 - val_loss: 0.0703
Epoch 23/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 500us/step - accuracy: 0.9769 - loss: 0.0879 - val_accuracy: 0.9840 - val_loss: 0.0689
Epoch 24/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 495us/step - accuracy: 0.9779 - loss: 0.0834 - val_accuracy: 0.9852 - val_loss: 0.0680
Epoch 25/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 497us/step - accuracy: 0.9776 - loss: 0.0832 - val_accuracy: 0.9855 - val_loss: 0.0684
Epoch 26/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9778 - loss: 0.0822 - val_accuracy: 0.9850 - val_loss: 0.0680
Epoch 27/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 496us/step - accuracy: 0.9779 - loss: 0.0821 - val_accuracy: 0.9858 - val_loss: 0.0680
Epoch 28/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9789 - loss: 0.0806 - val_accuracy: 0.9860 - val_loss: 0.0676
Epoch 29/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 503us/step - accuracy: 0.9789 - loss: 0.0806 - val_accuracy: 0.9858 - val_loss: 0.0661
Epoch 30/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 503us/step - accuracy: 0.9792 - loss: 0.0815 - val_accuracy: 0.9865 - val_loss: 0.0663
Epoch 31/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9806 - loss: 0.0778 - val_accuracy: 0.9862 - val_loss: 0.0652
Epoch 32/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9794 - loss: 0.0779 - val_accuracy: 0.9858 - val_loss: 0.0655
Epoch 33/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 496us/step - accuracy: 0.9799 - loss: 0.0755 - val_accuracy: 0.9865 - val_loss: 0.0650
Epoch 34/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 517us/step - accuracy: 0.9803 - loss: 0.0744 - val_accuracy: 0.9862 - val_loss: 0.0648
Epoch 35/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9809 - loss: 0.0755 - val_accuracy: 0.9872 - val_loss: 0.0641
Epoch 36/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 502us/step - accuracy: 0.9802 - loss: 0.0776 - val_accuracy: 0.9872 - val_loss: 0.0639
Epoch 37/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9801 - loss: 0.0744 - val_accuracy: 0.9877 - val_loss: 0.0638
Epoch 38/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9812 - loss: 0.0731 - val_accuracy: 0.9875 - val_loss: 0.0636
Epoch 39/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 580us/step - accuracy: 0.9818 - loss: 0.0714 - val_accuracy: 0.9883 - val_loss: 0.0632
Epoch 40/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 500us/step - accuracy: 0.9811 - loss: 0.0724 - val_accuracy: 0.9883 - val_loss: 0.0630
Epoch 41/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9806 - loss: 0.0756 - val_accuracy: 0.9880 - val_loss: 0.0629
Epoch 42/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 502us/step - accuracy: 0.9814 - loss: 0.0726 - val_accuracy: 0.9880 - val_loss: 0.0631
Epoch 43/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 499us/step - accuracy: 0.9814 - loss: 0.0721 - val_accuracy: 0.9883 - val_loss: 0.0622
Epoch 44/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 500us/step - accuracy: 0.9818 - loss: 0.0708 - val_accuracy: 0.9890 - val_loss: 0.0624
Epoch 45/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 500us/step - accuracy: 0.9831 - loss: 0.0680 - val_accuracy: 0.9890 - val_loss: 0.0617
Epoch 46/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 502us/step - accuracy: 0.9822 - loss: 0.0679 - val_accuracy: 0.9885 - val_loss: 0.0614
Epoch 47/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 496us/step - accuracy: 0.9823 - loss: 0.0674 - val_accuracy: 0.9887 - val_loss: 0.0609
Epoch 48/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 596us/step - accuracy: 0.9815 - loss: 0.0702 - val_accuracy: 0.9885 - val_loss: 0.0610
Epoch 49/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 505us/step - accuracy: 0.9834 - loss: 0.0673 - val_accuracy: 0.9890 - val_loss: 0.0612
Epoch 50/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 501us/step - accuracy: 0.9836 - loss: 0.0652 - val_accuracy: 0.9883 - val_loss: 0.0611
Time taken: 13.8s

plot_history(history_2, "loss")

perf_train["Model 2"] = model_performance_classification(model_2, X_train, y_train)
perf_val["Model 2"] = model_performance_classification(model_2, X_val, y_val)
print("Train:"); display(perf_train["Model 2"])
print("Validation:"); display(perf_val["Model 2"])

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 241us/step
125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 294us/step
Train:

Validation:

print_classification_report(model_2, X_train, y_train, "Train", "Model_2")
print_classification_report(model_2, X_val, y_val, "Validation", "Model_2")

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 239us/step
Classification Report - Train data Model_2

              precision    recall  f1-score   support

         0.0       0.99      1.00      0.99     15112
         1.0       0.98      0.82      0.89       888

    accuracy                           0.99     16000
   macro avg       0.99      0.91      0.94     16000
weighted avg       0.99      0.99      0.99     16000

125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 276us/step
Classification Report - Validation data Model_2

              precision    recall  f1-score   support

         0.0       0.99      1.00      0.99      3778
         1.0       0.98      0.81      0.88       222

    accuracy                           0.99      4000
   macro avg       0.98      0.90      0.94      4000
weighted avg       0.99      0.99      0.99      4000

config_3 = {"layers": [32, 16, 8], "activation": "relu", "dropout_rate": 0.5}
model_3 = build_model(X_train.shape[1], config_3, "sgd")
model_3.summary()

Model: "sequential"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ dense (Dense)                   │ (None, 32)             │         1,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dropout (Dropout)               │ (None, 32)             │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 16)             │           528 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_2 (Dense)                 │ (None, 8)              │           136 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_3 (Dense)                 │ (None, 1)              │             9 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 1,985 (7.75 KB)

 Trainable params: 1,985 (7.75 KB)

 Non-trainable params: 0 (0.00 B)

start = time.time()
history_3 = model_3.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    batch_size=BATCH_SIZE,
    epochs=EPOCHS,
    class_weight=cw_dict,
)
print(f"Time taken: {time.time() - start:.1f}s")

Epoch 1/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 1s 655us/step - accuracy: 0.7728 - loss: 0.9619 - val_accuracy: 0.9388 - val_loss: 0.2194
Epoch 2/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.8801 - loss: 0.7312 - val_accuracy: 0.9492 - val_loss: 0.2181
Epoch 3/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 510us/step - accuracy: 0.9201 - loss: 0.6240 - val_accuracy: 0.9645 - val_loss: 0.1783
Epoch 4/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9294 - loss: 0.5776 - val_accuracy: 0.9695 - val_loss: 0.1585
Epoch 5/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 506us/step - accuracy: 0.9396 - loss: 0.5572 - val_accuracy: 0.9740 - val_loss: 0.1664
Epoch 6/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 512us/step - accuracy: 0.9477 - loss: 0.5195 - val_accuracy: 0.9770 - val_loss: 0.1544
Epoch 7/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 507us/step - accuracy: 0.9516 - loss: 0.5190 - val_accuracy: 0.9765 - val_loss: 0.1526
Epoch 8/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 507us/step - accuracy: 0.9532 - loss: 0.4850 - val_accuracy: 0.9835 - val_loss: 0.1305
Epoch 9/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 541us/step - accuracy: 0.9579 - loss: 0.4962 - val_accuracy: 0.9787 - val_loss: 0.1418
Epoch 10/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9557 - loss: 0.4774 - val_accuracy: 0.9833 - val_loss: 0.1364
Epoch 11/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 509us/step - accuracy: 0.9573 - loss: 0.4772 - val_accuracy: 0.9852 - val_loss: 0.1261
Epoch 12/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 621us/step - accuracy: 0.9609 - loss: 0.4735 - val_accuracy: 0.9855 - val_loss: 0.1307
Epoch 13/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 596us/step - accuracy: 0.9591 - loss: 0.4556 - val_accuracy: 0.9840 - val_loss: 0.1324
Epoch 14/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 661us/step - accuracy: 0.9611 - loss: 0.4511 - val_accuracy: 0.9858 - val_loss: 0.1235
Epoch 15/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 505us/step - accuracy: 0.9615 - loss: 0.4452 - val_accuracy: 0.9872 - val_loss: 0.1222
Epoch 16/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 506us/step - accuracy: 0.9659 - loss: 0.4322 - val_accuracy: 0.9850 - val_loss: 0.1245
Epoch 17/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 515us/step - accuracy: 0.9608 - loss: 0.4534 - val_accuracy: 0.9865 - val_loss: 0.1349
Epoch 18/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 502us/step - accuracy: 0.9640 - loss: 0.4250 - val_accuracy: 0.9872 - val_loss: 0.1308
Epoch 19/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9665 - loss: 0.4263 - val_accuracy: 0.9868 - val_loss: 0.1184
Epoch 20/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 504us/step - accuracy: 0.9676 - loss: 0.4319 - val_accuracy: 0.9872 - val_loss: 0.1244
Epoch 21/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 507us/step - accuracy: 0.9683 - loss: 0.4356 - val_accuracy: 0.9868 - val_loss: 0.1185
Epoch 22/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 504us/step - accuracy: 0.9655 - loss: 0.4353 - val_accuracy: 0.9872 - val_loss: 0.1240
Epoch 23/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 511us/step - accuracy: 0.9681 - loss: 0.4184 - val_accuracy: 0.9883 - val_loss: 0.1191
Epoch 24/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 509us/step - accuracy: 0.9664 - loss: 0.4284 - val_accuracy: 0.9872 - val_loss: 0.1257
Epoch 25/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 511us/step - accuracy: 0.9690 - loss: 0.4177 - val_accuracy: 0.9850 - val_loss: 0.1266
Epoch 26/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9674 - loss: 0.4230 - val_accuracy: 0.9875 - val_loss: 0.1233
Epoch 27/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 504us/step - accuracy: 0.9677 - loss: 0.4167 - val_accuracy: 0.9868 - val_loss: 0.1259
Epoch 28/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 506us/step - accuracy: 0.9689 - loss: 0.4240 - val_accuracy: 0.9872 - val_loss: 0.1340
Epoch 29/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 607us/step - accuracy: 0.9684 - loss: 0.4273 - val_accuracy: 0.9858 - val_loss: 0.1323
Epoch 30/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9674 - loss: 0.4130 - val_accuracy: 0.9875 - val_loss: 0.1170
Epoch 31/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 516us/step - accuracy: 0.9678 - loss: 0.4261 - val_accuracy: 0.9860 - val_loss: 0.1267
Epoch 32/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 513us/step - accuracy: 0.9680 - loss: 0.4201 - val_accuracy: 0.9855 - val_loss: 0.1328
Epoch 33/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 525us/step - accuracy: 0.9680 - loss: 0.4240 - val_accuracy: 0.9860 - val_loss: 0.1257
Epoch 34/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 521us/step - accuracy: 0.9674 - loss: 0.4249 - val_accuracy: 0.9850 - val_loss: 0.1310
Epoch 35/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9689 - loss: 0.4095 - val_accuracy: 0.9875 - val_loss: 0.1178
Epoch 36/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 507us/step - accuracy: 0.9674 - loss: 0.4300 - val_accuracy: 0.9860 - val_loss: 0.1287
Epoch 37/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 503us/step - accuracy: 0.9672 - loss: 0.4439 - val_accuracy: 0.9877 - val_loss: 0.1348
Epoch 38/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9691 - loss: 0.4129 - val_accuracy: 0.9875 - val_loss: 0.1177
Epoch 39/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9684 - loss: 0.4177 - val_accuracy: 0.9890 - val_loss: 0.1208
Epoch 40/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 506us/step - accuracy: 0.9704 - loss: 0.4100 - val_accuracy: 0.9865 - val_loss: 0.1255
Epoch 41/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 506us/step - accuracy: 0.9712 - loss: 0.4074 - val_accuracy: 0.9868 - val_loss: 0.1237
Epoch 42/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 774us/step - accuracy: 0.9681 - loss: 0.4178 - val_accuracy: 0.9880 - val_loss: 0.1230
Epoch 43/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9719 - loss: 0.4098 - val_accuracy: 0.9883 - val_loss: 0.1275
Epoch 44/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 506us/step - accuracy: 0.9719 - loss: 0.4121 - val_accuracy: 0.9885 - val_loss: 0.1308
Epoch 45/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9710 - loss: 0.4174 - val_accuracy: 0.9885 - val_loss: 0.1267
Epoch 46/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 503us/step - accuracy: 0.9706 - loss: 0.4028 - val_accuracy: 0.9875 - val_loss: 0.1237
Epoch 47/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 509us/step - accuracy: 0.9717 - loss: 0.4110 - val_accuracy: 0.9883 - val_loss: 0.1167
Epoch 48/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 509us/step - accuracy: 0.9711 - loss: 0.4037 - val_accuracy: 0.9877 - val_loss: 0.1285
Epoch 49/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 515us/step - accuracy: 0.9699 - loss: 0.4036 - val_accuracy: 0.9883 - val_loss: 0.1222
Epoch 50/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 508us/step - accuracy: 0.9712 - loss: 0.3996 - val_accuracy: 0.9877 - val_loss: 0.1210
Time taken: 14.0s

plot_history(history_3, "loss")

perf_train["Model 3"] = model_performance_classification(model_3, X_train, y_train)
perf_val["Model 3"] = model_performance_classification(model_3, X_val, y_val)
print("Train:"); display(perf_train["Model 3"])
print("Validation:"); display(perf_val["Model 3"])

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 241us/step
125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 286us/step
Train:

Validation:

print_classification_report(model_3, X_train, y_train, "Train", "Model_3")
print_classification_report(model_3, X_val, y_val, "Validation", "Model_3")

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 236us/step
Classification Report - Train data Model_3

              precision    recall  f1-score   support

         0.0       0.99      1.00      0.99     15112
         1.0       0.94      0.89      0.91       888

    accuracy                           0.99     16000
   macro avg       0.96      0.94      0.95     16000
weighted avg       0.99      0.99      0.99     16000

125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 279us/step
Classification Report - Validation data Model_3

              precision    recall  f1-score   support

         0.0       0.99      0.99      0.99      3778
         1.0       0.90      0.87      0.89       222

    accuracy                           0.99      4000
   macro avg       0.95      0.93      0.94      4000
weighted avg       0.99      0.99      0.99      4000

config_4 = {"layers": [14, 7], "activation": "relu"}
model_4 = build_model(X_train.shape[1], config_4, "adam")
model_4.summary()

Model: "sequential"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ dense (Dense)                   │ (None, 14)             │           574 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 7)              │           105 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_2 (Dense)                 │ (None, 1)              │             8 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 687 (2.68 KB)

 Trainable params: 687 (2.68 KB)

 Non-trainable params: 0 (0.00 B)

start = time.time()
history_4 = model_4.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    batch_size=BATCH_SIZE,
    epochs=EPOCHS,
)
print(f"Time taken: {time.time() - start:.1f}s")

Epoch 1/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 1s 670us/step - accuracy: 0.9344 - loss: 0.1776 - val_accuracy: 0.9795 - val_loss: 0.0860
Epoch 2/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 528us/step - accuracy: 0.9812 - loss: 0.0755 - val_accuracy: 0.9835 - val_loss: 0.0744
Epoch 3/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 530us/step - accuracy: 0.9851 - loss: 0.0653 - val_accuracy: 0.9862 - val_loss: 0.0689
Epoch 4/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 530us/step - accuracy: 0.9872 - loss: 0.0593 - val_accuracy: 0.9875 - val_loss: 0.0650
Epoch 5/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 527us/step - accuracy: 0.9886 - loss: 0.0547 - val_accuracy: 0.9877 - val_loss: 0.0614
Epoch 6/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 727us/step - accuracy: 0.9896 - loss: 0.0511 - val_accuracy: 0.9887 - val_loss: 0.0595
Epoch 7/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 663us/step - accuracy: 0.9904 - loss: 0.0485 - val_accuracy: 0.9893 - val_loss: 0.0583
Epoch 8/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 533us/step - accuracy: 0.9909 - loss: 0.0464 - val_accuracy: 0.9893 - val_loss: 0.0583
Epoch 9/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 911us/step - accuracy: 0.9909 - loss: 0.0451 - val_accuracy: 0.9893 - val_loss: 0.0582
Epoch 10/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 546us/step - accuracy: 0.9912 - loss: 0.0437 - val_accuracy: 0.9895 - val_loss: 0.0581
Epoch 11/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 712us/step - accuracy: 0.9919 - loss: 0.0430 - val_accuracy: 0.9887 - val_loss: 0.0586
Epoch 12/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 594us/step - accuracy: 0.9919 - loss: 0.0422 - val_accuracy: 0.9887 - val_loss: 0.0582
Epoch 13/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 1s 1ms/step - accuracy: 0.9919 - loss: 0.0417 - val_accuracy: 0.9887 - val_loss: 0.0579
Epoch 14/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 546us/step - accuracy: 0.9921 - loss: 0.0411 - val_accuracy: 0.9887 - val_loss: 0.0580
Epoch 15/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 537us/step - accuracy: 0.9925 - loss: 0.0409 - val_accuracy: 0.9883 - val_loss: 0.0580
Epoch 16/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 547us/step - accuracy: 0.9925 - loss: 0.0404 - val_accuracy: 0.9887 - val_loss: 0.0580
Epoch 17/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 531us/step - accuracy: 0.9925 - loss: 0.0400 - val_accuracy: 0.9890 - val_loss: 0.0586
Epoch 18/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 535us/step - accuracy: 0.9924 - loss: 0.0397 - val_accuracy: 0.9890 - val_loss: 0.0585
Epoch 19/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 547us/step - accuracy: 0.9924 - loss: 0.0394 - val_accuracy: 0.9890 - val_loss: 0.0583
Epoch 20/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 529us/step - accuracy: 0.9924 - loss: 0.0392 - val_accuracy: 0.9895 - val_loss: 0.0589
Epoch 21/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 527us/step - accuracy: 0.9924 - loss: 0.0388 - val_accuracy: 0.9898 - val_loss: 0.0589
Epoch 22/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 526us/step - accuracy: 0.9927 - loss: 0.0386 - val_accuracy: 0.9898 - val_loss: 0.0592
Epoch 23/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 554us/step - accuracy: 0.9928 - loss: 0.0383 - val_accuracy: 0.9898 - val_loss: 0.0588
Epoch 24/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 542us/step - accuracy: 0.9927 - loss: 0.0379 - val_accuracy: 0.9898 - val_loss: 0.0588
Epoch 25/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 541us/step - accuracy: 0.9929 - loss: 0.0377 - val_accuracy: 0.9898 - val_loss: 0.0587
Epoch 26/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 571us/step - accuracy: 0.9929 - loss: 0.0373 - val_accuracy: 0.9895 - val_loss: 0.0594
Epoch 27/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 550us/step - accuracy: 0.9931 - loss: 0.0372 - val_accuracy: 0.9902 - val_loss: 0.0593
Epoch 28/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 545us/step - accuracy: 0.9927 - loss: 0.0370 - val_accuracy: 0.9902 - val_loss: 0.0593
Epoch 29/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 569us/step - accuracy: 0.9928 - loss: 0.0367 - val_accuracy: 0.9902 - val_loss: 0.0595
Epoch 30/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 725us/step - accuracy: 0.9931 - loss: 0.0367 - val_accuracy: 0.9910 - val_loss: 0.0590
Epoch 31/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 548us/step - accuracy: 0.9930 - loss: 0.0363 - val_accuracy: 0.9905 - val_loss: 0.0594
Epoch 32/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 541us/step - accuracy: 0.9930 - loss: 0.0361 - val_accuracy: 0.9910 - val_loss: 0.0593
Epoch 33/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 541us/step - accuracy: 0.9931 - loss: 0.0361 - val_accuracy: 0.9910 - val_loss: 0.0593
Epoch 34/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 539us/step - accuracy: 0.9931 - loss: 0.0360 - val_accuracy: 0.9908 - val_loss: 0.0589
Epoch 35/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 583us/step - accuracy: 0.9933 - loss: 0.0357 - val_accuracy: 0.9908 - val_loss: 0.0597
Epoch 36/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 604us/step - accuracy: 0.9933 - loss: 0.0355 - val_accuracy: 0.9905 - val_loss: 0.0598
Epoch 37/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 540us/step - accuracy: 0.9933 - loss: 0.0354 - val_accuracy: 0.9908 - val_loss: 0.0596
Epoch 38/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 529us/step - accuracy: 0.9933 - loss: 0.0353 - val_accuracy: 0.9905 - val_loss: 0.0593
Epoch 39/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 888us/step - accuracy: 0.9935 - loss: 0.0352 - val_accuracy: 0.9905 - val_loss: 0.0591
Epoch 40/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 716us/step - accuracy: 0.9932 - loss: 0.0352 - val_accuracy: 0.9905 - val_loss: 0.0594
Epoch 41/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 535us/step - accuracy: 0.9933 - loss: 0.0350 - val_accuracy: 0.9908 - val_loss: 0.0595
Epoch 42/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 541us/step - accuracy: 0.9933 - loss: 0.0349 - val_accuracy: 0.9905 - val_loss: 0.0598
Epoch 43/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 531us/step - accuracy: 0.9932 - loss: 0.0349 - val_accuracy: 0.9905 - val_loss: 0.0598
Epoch 44/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 584us/step - accuracy: 0.9933 - loss: 0.0348 - val_accuracy: 0.9908 - val_loss: 0.0599
Epoch 45/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 556us/step - accuracy: 0.9934 - loss: 0.0347 - val_accuracy: 0.9908 - val_loss: 0.0600
Epoch 46/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 565us/step - accuracy: 0.9935 - loss: 0.0345 - val_accuracy: 0.9910 - val_loss: 0.0604
Epoch 47/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 910us/step - accuracy: 0.9931 - loss: 0.0343 - val_accuracy: 0.9905 - val_loss: 0.0609
Epoch 48/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 616us/step - accuracy: 0.9933 - loss: 0.0344 - val_accuracy: 0.9910 - val_loss: 0.0609
Epoch 49/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 662us/step - accuracy: 0.9933 - loss: 0.0342 - val_accuracy: 0.9905 - val_loss: 0.0611
Epoch 50/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 582us/step - accuracy: 0.9932 - loss: 0.0341 - val_accuracy: 0.9912 - val_loss: 0.0614
Time taken: 16.0s

plot_history(history_4, "loss")

perf_train["Model 4"] = model_performance_classification(model_4, X_train, y_train)
perf_val["Model 4"] = model_performance_classification(model_4, X_val, y_val)
print("Train:"); display(perf_train["Model 4"])
print("Validation:"); display(perf_val["Model 4"])

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 493us/step
125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 444us/step
Train:

Validation:

print_classification_report(model_4, X_train, y_train, "Train", "Model_4")
print_classification_report(model_4, X_val, y_val, "Validation", "Model_4")

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 246us/step
Classification Report - Train data Model_4

              precision    recall  f1-score   support

         0.0       0.99      1.00      1.00     15112
         1.0       0.99      0.89      0.94       888

    accuracy                           0.99     16000
   macro avg       0.99      0.94      0.97     16000
weighted avg       0.99      0.99      0.99     16000

125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 298us/step
Classification Report - Validation data Model_4

              precision    recall  f1-score   support

         0.0       0.99      1.00      1.00      3778
         1.0       0.98      0.86      0.92       222

    accuracy                           0.99      4000
   macro avg       0.99      0.93      0.96      4000
weighted avg       0.99      0.99      0.99      4000

config_5 = {"layers": [32, 16, 8], "activation": "relu", "dropout_rate": 0.5}
model_5 = build_model(X_train.shape[1], config_5, "adam")
model_5.summary()

Model: "sequential"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ dense (Dense)                   │ (None, 32)             │         1,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dropout (Dropout)               │ (None, 32)             │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 16)             │           528 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_2 (Dense)                 │ (None, 8)              │           136 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_3 (Dense)                 │ (None, 1)              │             9 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 1,985 (7.75 KB)

 Trainable params: 1,985 (7.75 KB)

 Non-trainable params: 0 (0.00 B)

start = time.time()
history_5 = model_5.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    batch_size=BATCH_SIZE,
    epochs=EPOCHS,
)
print(f"Time taken: {time.time() - start:.1f}s")

Epoch 1/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 1s 729us/step - accuracy: 0.8983 - loss: 0.3573 - val_accuracy: 0.9665 - val_loss: 0.1213
Epoch 2/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 580us/step - accuracy: 0.9604 - loss: 0.1368 - val_accuracy: 0.9740 - val_loss: 0.0974
Epoch 3/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 579us/step - accuracy: 0.9684 - loss: 0.1128 - val_accuracy: 0.9805 - val_loss: 0.0860
Epoch 4/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 591us/step - accuracy: 0.9711 - loss: 0.1032 - val_accuracy: 0.9837 - val_loss: 0.0786
Epoch 5/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 582us/step - accuracy: 0.9752 - loss: 0.0939 - val_accuracy: 0.9858 - val_loss: 0.0719
Epoch 6/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 596us/step - accuracy: 0.9776 - loss: 0.0887 - val_accuracy: 0.9875 - val_loss: 0.0689
Epoch 7/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 581us/step - accuracy: 0.9779 - loss: 0.0856 - val_accuracy: 0.9868 - val_loss: 0.0679
Epoch 8/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 805us/step - accuracy: 0.9804 - loss: 0.0783 - val_accuracy: 0.9890 - val_loss: 0.0652
Epoch 9/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 601us/step - accuracy: 0.9803 - loss: 0.0786 - val_accuracy: 0.9887 - val_loss: 0.0652
Epoch 10/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 725us/step - accuracy: 0.9820 - loss: 0.0763 - val_accuracy: 0.9880 - val_loss: 0.0661
Epoch 11/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 698us/step - accuracy: 0.9816 - loss: 0.0733 - val_accuracy: 0.9883 - val_loss: 0.0635
Epoch 12/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 634us/step - accuracy: 0.9810 - loss: 0.0755 - val_accuracy: 0.9893 - val_loss: 0.0637
Epoch 13/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 589us/step - accuracy: 0.9826 - loss: 0.0726 - val_accuracy: 0.9895 - val_loss: 0.0623
Epoch 14/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 600us/step - accuracy: 0.9824 - loss: 0.0730 - val_accuracy: 0.9895 - val_loss: 0.0626
Epoch 15/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 605us/step - accuracy: 0.9826 - loss: 0.0724 - val_accuracy: 0.9893 - val_loss: 0.0613
Epoch 16/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 620us/step - accuracy: 0.9833 - loss: 0.0684 - val_accuracy: 0.9898 - val_loss: 0.0607
Epoch 17/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 619us/step - accuracy: 0.9849 - loss: 0.0651 - val_accuracy: 0.9900 - val_loss: 0.0605
Epoch 18/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 726us/step - accuracy: 0.9836 - loss: 0.0684 - val_accuracy: 0.9905 - val_loss: 0.0596
Epoch 19/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 603us/step - accuracy: 0.9844 - loss: 0.0646 - val_accuracy: 0.9900 - val_loss: 0.0581
Epoch 20/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 602us/step - accuracy: 0.9859 - loss: 0.0644 - val_accuracy: 0.9908 - val_loss: 0.0587
Epoch 21/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 590us/step - accuracy: 0.9857 - loss: 0.0634 - val_accuracy: 0.9900 - val_loss: 0.0623
Epoch 22/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 591us/step - accuracy: 0.9854 - loss: 0.0632 - val_accuracy: 0.9900 - val_loss: 0.0594
Epoch 23/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 613us/step - accuracy: 0.9864 - loss: 0.0608 - val_accuracy: 0.9898 - val_loss: 0.0592
Epoch 24/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 671us/step - accuracy: 0.9858 - loss: 0.0607 - val_accuracy: 0.9908 - val_loss: 0.0577
Epoch 25/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 594us/step - accuracy: 0.9856 - loss: 0.0615 - val_accuracy: 0.9905 - val_loss: 0.0584
Epoch 26/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 767us/step - accuracy: 0.9864 - loss: 0.0595 - val_accuracy: 0.9908 - val_loss: 0.0579
Epoch 27/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 598us/step - accuracy: 0.9862 - loss: 0.0601 - val_accuracy: 0.9902 - val_loss: 0.0549
Epoch 28/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 757us/step - accuracy: 0.9872 - loss: 0.0573 - val_accuracy: 0.9910 - val_loss: 0.0552
Epoch 29/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 624us/step - accuracy: 0.9861 - loss: 0.0594 - val_accuracy: 0.9910 - val_loss: 0.0566
Epoch 30/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 606us/step - accuracy: 0.9865 - loss: 0.0589 - val_accuracy: 0.9902 - val_loss: 0.0567
Epoch 31/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 600us/step - accuracy: 0.9854 - loss: 0.0595 - val_accuracy: 0.9898 - val_loss: 0.0566
Epoch 32/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 615us/step - accuracy: 0.9869 - loss: 0.0564 - val_accuracy: 0.9910 - val_loss: 0.0569
Epoch 33/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 595us/step - accuracy: 0.9864 - loss: 0.0582 - val_accuracy: 0.9908 - val_loss: 0.0554
Epoch 34/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 601us/step - accuracy: 0.9868 - loss: 0.0549 - val_accuracy: 0.9908 - val_loss: 0.0563
Epoch 35/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 598us/step - accuracy: 0.9866 - loss: 0.0590 - val_accuracy: 0.9900 - val_loss: 0.0573
Epoch 36/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 593us/step - accuracy: 0.9864 - loss: 0.0577 - val_accuracy: 0.9912 - val_loss: 0.0547
Epoch 37/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 594us/step - accuracy: 0.9868 - loss: 0.0562 - val_accuracy: 0.9908 - val_loss: 0.0557
Epoch 38/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 737us/step - accuracy: 0.9877 - loss: 0.0546 - val_accuracy: 0.9905 - val_loss: 0.0550
Epoch 39/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 1s 1ms/step - accuracy: 0.9877 - loss: 0.0548 - val_accuracy: 0.9908 - val_loss: 0.0540
Epoch 40/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 602us/step - accuracy: 0.9878 - loss: 0.0565 - val_accuracy: 0.9908 - val_loss: 0.0545
Epoch 41/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 608us/step - accuracy: 0.9874 - loss: 0.0558 - val_accuracy: 0.9915 - val_loss: 0.0541
Epoch 42/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 641us/step - accuracy: 0.9871 - loss: 0.0560 - val_accuracy: 0.9912 - val_loss: 0.0543
Epoch 43/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 599us/step - accuracy: 0.9876 - loss: 0.0558 - val_accuracy: 0.9908 - val_loss: 0.0559
Epoch 44/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 597us/step - accuracy: 0.9890 - loss: 0.0514 - val_accuracy: 0.9902 - val_loss: 0.0558
Epoch 45/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 592us/step - accuracy: 0.9880 - loss: 0.0534 - val_accuracy: 0.9912 - val_loss: 0.0560
Epoch 46/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 597us/step - accuracy: 0.9874 - loss: 0.0539 - val_accuracy: 0.9908 - val_loss: 0.0563
Epoch 47/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 614us/step - accuracy: 0.9883 - loss: 0.0545 - val_accuracy: 0.9910 - val_loss: 0.0550
Epoch 48/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 595us/step - accuracy: 0.9877 - loss: 0.0531 - val_accuracy: 0.9908 - val_loss: 0.0552
Epoch 49/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 600us/step - accuracy: 0.9888 - loss: 0.0531 - val_accuracy: 0.9902 - val_loss: 0.0539
Epoch 50/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 722us/step - accuracy: 0.9882 - loss: 0.0536 - val_accuracy: 0.9910 - val_loss: 0.0555
Time taken: 17.0s

plot_history(history_5, "loss")

perf_train["Model 5"] = model_performance_classification(model_5, X_train, y_train)
perf_val["Model 5"] = model_performance_classification(model_5, X_val, y_val)
print("Train:"); display(perf_train["Model 5"])
print("Validation:"); display(perf_val["Model 5"])

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 241us/step
125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 310us/step
Train:

Validation:

print_classification_report(model_5, X_train, y_train, "Train", "Model_5")
print_classification_report(model_5, X_val, y_val, "Validation", "Model_5")

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 232us/step
Classification Report - Train data Model_5

              precision    recall  f1-score   support

         0.0       0.99      1.00      1.00     15112
         1.0       0.99      0.88      0.93       888

    accuracy                           0.99     16000
   macro avg       0.99      0.94      0.96     16000
weighted avg       0.99      0.99      0.99     16000

125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 283us/step
Classification Report - Validation data Model_5

              precision    recall  f1-score   support

         0.0       0.99      1.00      1.00      3778
         1.0       0.97      0.86      0.91       222

    accuracy                           0.99      4000
   macro avg       0.98      0.93      0.95      4000
weighted avg       0.99      0.99      0.99      4000

config_6 = {"layers": [32, 16, 8], "activation": "relu", "dropout_rate": 0.5}
model_6 = build_model(X_train.shape[1], config_6, "sgd")
model_6.summary()

Model: "sequential"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ dense (Dense)                   │ (None, 32)             │         1,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dropout (Dropout)               │ (None, 32)             │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 16)             │           528 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_2 (Dense)                 │ (None, 8)              │           136 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_3 (Dense)                 │ (None, 1)              │             9 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 1,985 (7.75 KB)

 Trainable params: 1,985 (7.75 KB)

 Non-trainable params: 0 (0.00 B)

start = time.time()
history_6 = model_6.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    batch_size=BATCH_SIZE,
    epochs=EPOCHS,
    class_weight=cw_dict,
)
print(f"Time taken: {time.time() - start:.1f}s")

Epoch 1/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 1s 796us/step - accuracy: 0.7305 - loss: 1.0696 - val_accuracy: 0.8975 - val_loss: 0.3037
Epoch 2/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 516us/step - accuracy: 0.8751 - loss: 0.7323 - val_accuracy: 0.9335 - val_loss: 0.2401
Epoch 3/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 527us/step - accuracy: 0.8957 - loss: 0.6624 - val_accuracy: 0.9507 - val_loss: 0.2342
Epoch 4/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 575us/step - accuracy: 0.9154 - loss: 0.6236 - val_accuracy: 0.9635 - val_loss: 0.2023
Epoch 5/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 545us/step - accuracy: 0.9345 - loss: 0.5782 - val_accuracy: 0.9697 - val_loss: 0.1904
Epoch 6/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 618us/step - accuracy: 0.9396 - loss: 0.5602 - val_accuracy: 0.9735 - val_loss: 0.1634
Epoch 7/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 526us/step - accuracy: 0.9452 - loss: 0.5496 - val_accuracy: 0.9700 - val_loss: 0.1942
Epoch 8/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 790us/step - accuracy: 0.9474 - loss: 0.5220 - val_accuracy: 0.9643 - val_loss: 0.1931
Epoch 9/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 545us/step - accuracy: 0.9524 - loss: 0.5106 - val_accuracy: 0.9758 - val_loss: 0.1754
Epoch 10/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 525us/step - accuracy: 0.9482 - loss: 0.5103 - val_accuracy: 0.9772 - val_loss: 0.1530
Epoch 11/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 545us/step - accuracy: 0.9561 - loss: 0.4910 - val_accuracy: 0.9808 - val_loss: 0.1693
Epoch 12/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 579us/step - accuracy: 0.9567 - loss: 0.4707 - val_accuracy: 0.9780 - val_loss: 0.1718
Epoch 13/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 526us/step - accuracy: 0.9547 - loss: 0.4984 - val_accuracy: 0.9815 - val_loss: 0.1502
Epoch 14/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 511us/step - accuracy: 0.9588 - loss: 0.4710 - val_accuracy: 0.9783 - val_loss: 0.1555
Epoch 15/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 517us/step - accuracy: 0.9578 - loss: 0.4814 - val_accuracy: 0.9818 - val_loss: 0.1528
Epoch 16/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 518us/step - accuracy: 0.9584 - loss: 0.4599 - val_accuracy: 0.9787 - val_loss: 0.1501
Epoch 17/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 515us/step - accuracy: 0.9548 - loss: 0.4940 - val_accuracy: 0.9800 - val_loss: 0.1425
Epoch 18/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 531us/step - accuracy: 0.9563 - loss: 0.4739 - val_accuracy: 0.9812 - val_loss: 0.1427
Epoch 19/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 679us/step - accuracy: 0.9624 - loss: 0.4519 - val_accuracy: 0.9840 - val_loss: 0.1374
Epoch 20/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 518us/step - accuracy: 0.9624 - loss: 0.4462 - val_accuracy: 0.9843 - val_loss: 0.1375
Epoch 21/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 510us/step - accuracy: 0.9660 - loss: 0.4316 - val_accuracy: 0.9822 - val_loss: 0.1376
Epoch 22/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 515us/step - accuracy: 0.9626 - loss: 0.4499 - val_accuracy: 0.9852 - val_loss: 0.1242
Epoch 23/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 514us/step - accuracy: 0.9632 - loss: 0.4532 - val_accuracy: 0.9845 - val_loss: 0.1283
Epoch 24/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 544us/step - accuracy: 0.9610 - loss: 0.4520 - val_accuracy: 0.9850 - val_loss: 0.1213
Epoch 25/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 524us/step - accuracy: 0.9634 - loss: 0.4326 - val_accuracy: 0.9868 - val_loss: 0.1223
Epoch 26/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 550us/step - accuracy: 0.9661 - loss: 0.4458 - val_accuracy: 0.9847 - val_loss: 0.1301
Epoch 27/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 575us/step - accuracy: 0.9662 - loss: 0.4269 - val_accuracy: 0.9865 - val_loss: 0.1206
Epoch 28/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 557us/step - accuracy: 0.9660 - loss: 0.4324 - val_accuracy: 0.9847 - val_loss: 0.1361
Epoch 29/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 533us/step - accuracy: 0.9639 - loss: 0.4213 - val_accuracy: 0.9885 - val_loss: 0.1158
Epoch 30/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 531us/step - accuracy: 0.9681 - loss: 0.4273 - val_accuracy: 0.9875 - val_loss: 0.1267
Epoch 31/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 671us/step - accuracy: 0.9678 - loss: 0.4159 - val_accuracy: 0.9860 - val_loss: 0.1242
Epoch 32/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 534us/step - accuracy: 0.9659 - loss: 0.4322 - val_accuracy: 0.9852 - val_loss: 0.1349
Epoch 33/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 566us/step - accuracy: 0.9685 - loss: 0.4368 - val_accuracy: 0.9862 - val_loss: 0.1146
Epoch 34/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 582us/step - accuracy: 0.9673 - loss: 0.4240 - val_accuracy: 0.9858 - val_loss: 0.1192
Epoch 35/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 567us/step - accuracy: 0.9642 - loss: 0.4355 - val_accuracy: 0.9845 - val_loss: 0.1282
Epoch 36/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 549us/step - accuracy: 0.9651 - loss: 0.4205 - val_accuracy: 0.9860 - val_loss: 0.1217
Epoch 37/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 535us/step - accuracy: 0.9676 - loss: 0.4212 - val_accuracy: 0.9877 - val_loss: 0.1215
Epoch 38/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 559us/step - accuracy: 0.9678 - loss: 0.4160 - val_accuracy: 0.9880 - val_loss: 0.1196
Epoch 39/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 531us/step - accuracy: 0.9676 - loss: 0.4184 - val_accuracy: 0.9883 - val_loss: 0.1252
Epoch 40/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 694us/step - accuracy: 0.9654 - loss: 0.4265 - val_accuracy: 0.9860 - val_loss: 0.1279
Epoch 41/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 532us/step - accuracy: 0.9678 - loss: 0.4127 - val_accuracy: 0.9860 - val_loss: 0.1271
Epoch 42/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 536us/step - accuracy: 0.9669 - loss: 0.4383 - val_accuracy: 0.9847 - val_loss: 0.1285
Epoch 43/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 552us/step - accuracy: 0.9693 - loss: 0.4161 - val_accuracy: 0.9870 - val_loss: 0.1192
Epoch 44/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 526us/step - accuracy: 0.9672 - loss: 0.4159 - val_accuracy: 0.9885 - val_loss: 0.1154
Epoch 45/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 536us/step - accuracy: 0.9697 - loss: 0.4166 - val_accuracy: 0.9880 - val_loss: 0.1250
Epoch 46/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 537us/step - accuracy: 0.9709 - loss: 0.4160 - val_accuracy: 0.9885 - val_loss: 0.1149
Epoch 47/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 537us/step - accuracy: 0.9697 - loss: 0.4121 - val_accuracy: 0.9883 - val_loss: 0.1153
Epoch 48/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 567us/step - accuracy: 0.9706 - loss: 0.4051 - val_accuracy: 0.9862 - val_loss: 0.1192
Epoch 49/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 659us/step - accuracy: 0.9681 - loss: 0.4127 - val_accuracy: 0.9870 - val_loss: 0.1184
Epoch 50/50
500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 555us/step - accuracy: 0.9674 - loss: 0.4222 - val_accuracy: 0.9890 - val_loss: 0.1107
Time taken: 14.8s

plot_history(history_6, "loss")

perf_train["Model 6"] = model_performance_classification(model_6, X_train, y_train)
perf_val["Model 6"] = model_performance_classification(model_6, X_val, y_val)
print("Train:"); display(perf_train["Model 6"])
print("Validation:"); display(perf_val["Model 6"])

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 239us/step
125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 278us/step
Train:

Validation:

print_classification_report(model_6, X_train, y_train, "Train", "Model_6")
print_classification_report(model_6, X_val, y_val, "Validation", "Model_6")

500/500 ━━━━━━━━━━━━━━━━━━━━ 0s 229us/step
Classification Report - Train data Model_6

              precision    recall  f1-score   support

         0.0       0.99      1.00      1.00     15112
         1.0       0.92      0.91      0.91       888

    accuracy                           0.99     16000
   macro avg       0.96      0.95      0.95     16000
weighted avg       0.99      0.99      0.99     16000

125/125 ━━━━━━━━━━━━━━━━━━━━ 0s 275us/step
Classification Report - Validation data Model_6

              precision    recall  f1-score   support

         0.0       0.99      1.00      0.99      3778
         1.0       0.92      0.87      0.90       222

    accuracy                           0.99      4000
   macro avg       0.96      0.93      0.95      4000
weighted avg       0.99      0.99      0.99      4000

train_comp, val_comp = compare_models(perf_train, perf_val)

print("Training set performance comparison:")
display(train_comp)
print()
print("Validation set performance comparison:")
display(val_comp)

Training set performance comparison:

Validation set performance comparison:

# TODO: Uncomment the best model after reviewing the comparison table
# best_model = model_0
# best_model = model_1
# best_model = model_2
# best_model = model_3
best_model = model_4
# best_model = model_5
# best_model = model_6

best_model_name = [name for name, val in globals().items() if val is best_model][0]
print(f"Best model: {best_model_name}")

Best model: model_4

import numpy as np
import matplotlib.pyplot as plt
from matplotlib.collections import LineCollection
from matplotlib.patches import Circle, Patch
from IPython.display import Markdown, display


def _select_indices(n_units, max_display):
    if n_units <= max_display:
        return np.arange(n_units)
    return np.unique(np.linspace(0, n_units - 1, max_display, dtype=int))


def _y_positions(count):
    if count == 1:
        return np.array([0.5])
    return np.linspace(0.92, 0.08, count)


def visualize_neural_network(model, model_label, max_neurons_per_layer=40):
    # Build dense-layer structure with trained weight matrices
    dense_layers = [layer for layer in model.layers if layer.__class__.__name__ == "Dense"]
    input_units = int(model.input_shape[-1])

    layer_sizes = [input_units] + [int(layer.units) for layer in dense_layers]
    layer_names = ["Input"] + [f"Dense {i}" for i in range(1, len(dense_layers))] + ["Output"]

    # x coordinates for layers
    x_positions = np.linspace(0.06, 0.94, len(layer_sizes))

    # displayed neurons and coordinates
    selected = [_select_indices(n, max_neurons_per_layer) for n in layer_sizes]
    node_coords = []
    for x, idx in zip(x_positions, selected):
        y = _y_positions(len(idx))
        node_coords.append([(x, yi) for yi in y])

    # figure setup
    fig, ax = plt.subplots(figsize=(14, 8), dpi=150)
    ax.set_facecolor("#f7f9fc")
    fig.patch.set_facecolor("#eef2f7")

    # draw weighted connections between consecutive layers
    for li in range(1, len(layer_sizes)):
        w = dense_layers[li - 1].get_weights()[0]  # shape: (prev, curr)
        prev_idx = selected[li - 1]
        curr_idx = selected[li]

        segments = []
        weights = []
        for i_disp, i_real in enumerate(prev_idx):
            x1, y1 = node_coords[li - 1][i_disp]
            for j_disp, j_real in enumerate(curr_idx):
                x2, y2 = node_coords[li][j_disp]
                segments.append([(x1, y1), (x2, y2)])
                weights.append(w[i_real, j_real])

        weights = np.array(weights)
        max_abs = np.max(np.abs(weights)) if len(weights) else 1.0
        norm = np.abs(weights) / (max_abs + 1e-9)
        colors = plt.cm.coolwarm((weights + max_abs) / (2 * max_abs + 1e-9))
        colors[:, 3] = 0.08 + 0.55 * norm  # alpha based on strength

        lc = LineCollection(segments, colors=colors, linewidths=0.35 + 1.2 * norm, zorder=1)
        ax.add_collection(lc)

    # draw neurons
    input_color = "#4da3ff"
    hidden_color = "#ffad5a"
    output_color = "#5ccf8d"

    for li, coords in enumerate(node_coords):
        if li == 0:
            color = input_color
        elif li == len(node_coords) - 1:
            color = output_color
        else:
            color = hidden_color

        for x, y in coords:
            ax.add_patch(Circle((x, y), radius=0.0115, facecolor=color, edgecolor="#1f2937", linewidth=0.5, zorder=3))

    # layer titles and neuron counts
    for x, name, total, shown in zip(x_positions, layer_names, layer_sizes, selected):
        ax.text(x, 1.02, name, ha="center", va="bottom", fontsize=11, fontweight="bold")
        if len(shown) < total:
            ax.text(x, -0.02, f"showing {len(shown)}/{total}", ha="center", va="top", fontsize=9, color="#4b5563")
        else:
            ax.text(x, -0.02, f"{total} neurons", ha="center", va="top", fontsize=9, color="#4b5563")

    # metadata
    total_params = int(model.count_params())
    optimizer_name = model.optimizer.__class__.__name__
    ax.text(
        0.5,
        -0.1,
        f"{model_label} | Optimizer: {optimizer_name} | Parameters: {total_params:,}",
        ha="center",
        va="center",
        fontsize=10,
        color="#111827",
    )

    ax.set_title("Neural Network Topology (Neuron-Level View)", fontsize=16, pad=18, fontweight="bold")
    ax.set_xlim(0, 1)
    ax.set_ylim(0, 1)
    ax.axis("off")

    legend_items = [
        Patch(facecolor=input_color, edgecolor="#1f2937", label="Input Layer"),
        Patch(facecolor=hidden_color, edgecolor="#1f2937", label="Hidden Layer"),
        Patch(facecolor=output_color, edgecolor="#1f2937", label="Output Layer"),
    ]
    ax.legend(handles=legend_items, loc="upper center", bbox_to_anchor=(0.5, -0.16), ncol=3, frameon=False)

    plt.tight_layout()
    plt.show()


display(Markdown("#### Final Neural Network Architecture (Actual Neuron View)"))
visualize_neural_network(best_model, best_model_name, max_neurons_per_layer=40)

from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay

# Confusion matrices — validation and test
fig, axes = plt.subplots(1, 2, figsize=(12, 5))

for ax, (X_set, y_set, title) in zip(axes, [
    (X_val, y_val, "Validation Set"),
    (X_test, y_test, "Test Set"),
]):
    y_pred = (best_model.predict(X_set) > THRESHOLD).astype(int)
    cm = confusion_matrix(y_set, y_pred)
    disp = ConfusionMatrixDisplay(confusion_matrix=cm, display_labels=["Normal", "Failure"])
    disp.plot(ax=ax, cmap="Blues", values_format="d")
    ax.set_title(f"Confusion Matrix — {title}")

plt.tight_layout()
plt.show()

# Threshold sensitivity analysis
from sklearn.metrics import recall_score, precision_score, f1_score

thresholds = [0.3, 0.4, 0.5, 0.6, 0.7]
rows = []
y_prob_val = best_model.predict(X_val).ravel()

for t in thresholds:
    y_pred_t = (y_prob_val > t).astype(int)
    rows.append({
        "Threshold": t,
        "Recall (macro)": round(recall_score(y_val, y_pred_t, average="macro"), 4),
        "Precision (macro)": round(precision_score(y_val, y_pred_t, average="macro"), 4),
        "F1 (macro)": round(f1_score(y_val, y_pred_t, average="macro"), 4),
    })

threshold_df = pd.DataFrame(rows).set_index("Threshold")
print("Threshold Sensitivity — Validation Set:")
display(threshold_df)

best_model_test_perf = model_performance_classification(best_model, X_test, y_test)
print("Test set performance:")
display(best_model_test_perf)

157/157 ━━━━━━━━━━━━━━━━━━━━ 0s 505us/step
Test set performance:

print_classification_report(best_model, X_test, y_test, "Test", "Best Model")

157/157 ━━━━━━━━━━━━━━━━━━━━ 0s 405us/step
Classification Report - Test data Best Model

              precision    recall  f1-score   support

         0.0       0.99      1.00      0.99      4718
         1.0       0.97      0.85      0.90       282

    accuracy                           0.99      5000
   macro avg       0.98      0.92      0.95      5000
weighted avg       0.99      0.99      0.99      5000

Prediction Outcome	Operational Meaning	Cost Impact
True Positive (TP)	Failure correctly predicted	Repair cost (acceptable)
False Negative (FN)	Failure missed	Replacement cost (highest risk)
False Positive (FP)	Unnecessary alert	Inspection cost (lowest risk)

	V1	V2	V3	V4	V5	V6	V7	V8	V9	V10	V11	V12	V13	V14	V15	V16	V17	V18	V19	V20	V21	V22	V23	V24	V25	V26	V27	V28	V29	V30	V31	V32	V33	V34	V35	V36	V37	V38	V39	V40
0	-4.464606	-4.679129	3.101546	0.506130	-0.221083	-2.032511	-2.910870	0.050714	-1.522351	3.761892	-5.714719	0.735893	0.981251	1.417884	-3.375815	-3.047303	0.306194	2.914097	2.269979	4.394876	-2.388299	0.646388	-1.190508	3.132986	0.665277	-2.510846	-0.036744	0.726218	-3.982187	-1.072638	1.667098	3.059700	-1.690440	2.846296	2.235198	6.667486	0.443809	-2.369169	2.950578	-3.480324
1	3.365912	3.653381	0.909671	-1.367528	0.332016	2.358938	0.732600	-4.332135	0.565695	-0.101080	1.914465	-0.951458	-1.255259	-2.706522	0.193223	-4.769379	-2.205319	0.907716	0.756894	-5.833678	-3.065122	1.596647	-1.757311	1.766444	-0.267098	3.625036	1.500346	-0.585712	0.783034	-0.201217	0.024883	-1.795474	3.032780	-2.467514	1.894599	-2.297780	-1.731048	5.908837	-0.386345	0.616242
2	-3.831843	-5.824444	0.634031	-2.418815	-1.773827	1.016824	-2.098941	-3.173204	-2.081860	5.392621	-0.770673	1.106718	1.144261	0.943301	-3.163804	-4.247825	-4.038909	3.688534	3.311196	1.059002	-2.143026	1.650120	-1.660592	1.679910	-0.450782	-4.550695	3.738779	1.134404	-2.033531	0.840839	-1.600395	-0.257101	0.803550	4.086219	2.292138	5.360850	0.351993	2.940021	3.839160	-4.309402
3	1.618098	1.888342	7.046143	-1.147285	0.083080	-1.529780	0.207309	-2.493629	0.344926	2.118578	-3.053023	0.459719	2.704527	-0.636086	-0.453717	-3.174046	-3.404347	-1.281536	1.582104	-1.951778	-3.516555	-1.206011	-5.627854	-1.817653	2.124142	5.294642	4.748137	-2.308536	-3.962977	-6.028730	4.948770	-3.584425	-2.577474	1.363769	0.622714	5.550100	-1.526796	0.138853	3.101430	-1.277378
4	-0.111440	3.872488	-3.758361	-2.982897	3.792714	0.544960	0.205433	4.848994	-1.854920	-6.220023	1.998347	4.723757	0.709113	-1.989432	-2.632684	4.184447	2.245356	3.734452	-6.312766	-5.379918	-0.886667	2.061694	9.445586	4.489976	-3.945144	4.582065	-8.780422	-3.382967	5.106507	6.787513	2.044184	8.265896	6.629213	-10.068689	1.222987	-3.229763	1.686909	-2.163896	-3.644622	6.510338

	V1	V2	V3	V4	V5	V6	V7	V8	V9	V10	V11	V12	V13	V14	V15	V16	V17	V18	V19	V20	V21	V22	V23	V24	V25	V26	V27	V28	V29	V30	V31	V32	V33	V34	V35	V36	V37	V38	V39	V40
0	-0.613489	-3.819640	2.202302	1.300420	-1.184929	-4.495964	-1.835817	4.722989	1.206140	-0.341909	-5.122874	1.017021	4.818549	3.269001	-2.984330	1.387370	2.032002	-0.511587	-1.023069	7.338733	-2.242244	0.155489	2.053786	-2.772273	1.851369	-1.788696	-0.277282	-1.255143	-3.832886	-1.504542	1.586765	2.291204	-5.411388	0.870073	0.574479	4.157191	1.428093	-10.511342	0.454664	-1.448363
1	0.389608	-0.512341	0.527053	-2.576776	-1.016766	2.235112	-0.441301	-4.405744	-0.332869	1.966794	1.796544	0.410490	0.638328	-1.389600	-1.883410	-5.017922	-3.827238	2.418060	1.762285	-3.242297	-3.192960	1.857454	-1.707954	0.633444	-0.587898	0.083683	3.013935	-0.182309	0.223917	0.865228	-1.782158	-2.474936	2.493582	0.315165	2.059288	0.683859	-0.485452	5.128350	1.720744	-1.488235
2	-0.874861	-0.640632	4.084202	-1.590454	0.525855	-1.957592	-0.695367	1.347309	-1.732348	0.466500	-4.928214	3.565070	-0.449329	-0.656246	-0.166537	-1.630207	2.291865	2.396492	0.601278	1.793534	-2.120238	0.481968	-0.840707	1.790197	1.874395	0.363930	-0.169063	-0.483832	-2.118982	-2.156586	2.907291	-1.318888	-2.997464	0.459664	0.619774	5.631504	1.323512	-1.752154	1.808302	1.675748
3	0.238384	1.458607	4.014528	2.534478	1.196987	-3.117330	-0.924035	0.269493	1.322436	0.702345	-5.578345	-0.850662	2.590525	0.767418	-2.390809	-2.341961	0.571875	-0.933751	0.508677	1.210715	-3.259524	0.104587	-0.658875	1.498107	1.100305	4.142988	-0.248446	-1.136516	-5.355810	-4.545931	3.808667	3.517918	-3.074085	-0.284220	0.954576	3.029331	-1.367198	-3.412140	0.906000	-2.450889
4	5.828225	2.768260	-1.234530	2.809264	-1.641648	-1.406698	0.568643	0.965043	1.918379	-2.774855	-0.530016	1.374544	-0.650941	-1.679466	-0.379220	-4.443143	3.893857	-0.607640	2.944931	0.367233	-5.789081	4.597528	4.450264	3.224941	0.396701	0.247765	-2.362047	1.079378	-0.473076	2.242810	-3.591421	1.773841	-1.501573	-2.226702	4.776830	-6.559698	-0.805551	-0.276007	-3.858207	-0.537694

	V1	V2	V3	V4	V5	V6	V7	V8	V9	V10	V11	V12	V13	V14	V15	V16	V17	V18	V19	V20	V21	V22	V23	V24	V25	V26	V27	V28	V29	V30	V31	V32	V33	V34	V35	V36	V37	V38	V39	V40	Target
count	19982.000000	19982.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000	20000.000000
mean	-0.271996	0.440430	2.484699	-0.083152	-0.053752	-0.995443	-0.879325	-0.548195	-0.016808	-0.012998	-1.895393	1.604825	1.580486	-0.950632	-2.414993	-2.925225	-0.134261	1.189347	1.181808	0.023608	-3.611252	0.951835	-0.366116	1.134389	-0.002186	1.873785	-0.612413	-0.883218	-0.985625	-0.015534	0.486842	0.303799	0.049825	-0.462702	2.229620	1.514809	0.011316	-0.344025	0.890653	-0.875630	0.055500
std	3.441625	3.150784	3.388963	3.431595	2.104801	2.040970	1.761626	3.295756	2.160568	2.193201	3.124322	2.930454	2.874658	1.789651	3.354974	4.221717	3.345462	2.592276	3.396925	3.669477	3.567690	1.651547	4.031860	3.912069	2.016740	3.435137	4.368847	1.917713	2.684365	3.005258	3.461384	5.500400	3.575285	3.183841	2.937102	3.800860	1.788165	3.948147	1.753054	3.012155	0.228959
min	-11.876451	-12.319951	-10.708139	-15.082052	-8.603361	-10.227147	-7.949681	-15.657561	-8.596313	-9.853957	-14.832058	-12.948007	-13.228247	-7.738593	-16.416606	-20.374158	-14.091184	-11.643994	-13.491784	-13.922659	-17.956231	-10.122095	-14.866128	-16.387147	-8.228266	-11.834271	-14.904939	-9.269489	-12.579469	-14.796047	-13.722760	-19.876502	-16.898353	-17.985094	-15.349803	-14.833178	-5.478350	-17.375002	-6.438880	-11.023935	0.000000
25%	-2.737146	-1.640674	0.206860	-2.347660	-1.535607	-2.347238	-2.030926	-2.642665	-1.494973	-1.411212	-3.922404	-0.396514	-0.223545	-2.170741	-4.415322	-5.634240	-2.215611	-0.403917	-1.050168	-2.432953	-5.930360	-0.118127	-3.098756	-1.468062	-1.365178	-0.337863	-3.652323	-2.171218	-2.787443	-1.867114	-1.817772	-3.420469	-2.242857	-2.136984	0.336191	-0.943809	-1.255819	-2.987638	-0.272250	-2.940193	0.000000
50%	-0.747917	0.471536	2.255786	-0.135241	-0.101952	-1.000515	-0.917179	-0.389085	-0.067597	0.100973	-1.921237	1.507841	1.637185	-0.957163	-2.382617	-2.682705	-0.014580	0.883398	1.279061	0.033415	-3.532888	0.974687	-0.262093	0.969048	0.025050	1.950531	-0.884894	-0.891073	-1.176181	0.184346	0.490304	0.052073	-0.066249	-0.255008	2.098633	1.566526	-0.128435	-0.316849	0.919261	-0.920806	0.000000
75%	1.840112	2.543967	4.566165	2.130615	1.340480	0.380330	0.223695	1.722965	1.409203	1.477045	0.118906	3.571454	3.459886	0.270677	-0.359052	-0.095046	2.068751	2.571770	3.493299	2.512372	-1.265884	2.025594	2.451750	3.545975	1.397112	4.130037	2.189177	0.375884	0.629773	2.036229	2.730688	3.761722	2.255134	1.436935	4.064358	3.983939	1.175533	2.279399	2.057540	1.119897	0.000000
max	15.493002	13.089269	17.090919	13.236381	8.133797	6.975847	8.006091	11.679495	8.137580	8.108472	11.826433	15.080698	15.419616	5.670664	12.246455	13.583212	16.756432	13.179863	13.237742	16.052339	13.840473	7.409856	14.458734	17.163291	8.223389	16.836410	17.560404	6.527643	10.722055	12.505812	17.255090	23.633187	16.692486	14.358213	15.291065	19.329576	7.467006	15.289923	7.759877	10.654265	1.000000

Finding	Detail
Dataset size	20,000 training samples; 5,000 test samples; 40 sensor features
Class imbalance	Target = 1 (failure) represents ~8% of training data — roughly 12:1 ratio
Missing values	Present in both train and test; handled via median imputation
Feature distributions	Many features are right-skewed with outliers (sensor readings)
Correlations	Several feature pairs show moderate-to-high correlation (> 0.7), but no feature removal is applied to preserve information

ReneWind Predictive Maintenance — Full-Code Implementation¶

1. Business Problem Summary¶

2. Setup & Reproducibility¶

3. Data Loading & Overview¶

4. Exploratory Data Analysis¶

4.1 Univariate Analysis¶

4.2 Target Distribution¶

4.3 Correlation Heatmap¶

4.4 EDA Summary¶

5. Data Preprocessing¶

Summary of Improvement Methods Explored¶

6. Neural Network Models¶

Model 0 — Baseline (1 hidden layer, SGD, no dropout, no class weights)¶

Model 1 — Deeper Network (2 hidden layers, SGD, no dropout, no class weights)¶

Model 2 — Deeper + Dropout (3 hidden layers, SGD, dropout, no class weights)¶

Model 3 — Dropout + Class Weights (3 hidden layers, SGD, dropout, class weights)¶

Model 4 — Adam Optimizer (2 hidden layers, Adam, no dropout, no class weights)¶

Model 5 — Adam + Dropout (3 hidden layers, Adam, dropout, no class weights)¶

Model 6 — SGD + Dropout + Class Weights (3 hidden layers, SGD, dropout, class weights)¶

7. Performance Summary Table¶

8. Final Model Justification¶

Why Model 4?¶

Confusion Matrix & Threshold Analysis¶

Best Model Architecture Visualization¶

9. Test Set Evaluation¶

10. Business Recommendations¶

Deployment Recommendation¶

Operational Impact¶

Risk Considerations¶

Future Improvements¶

11. Non-Technical Executive Summary¶

What We Did¶

What We Discovered¶

What This Means for ReneWind¶

#	Method	Models Using It	Rationale
1	Increased network depth (1 → 2 → 3 hidden layers)	All models	Captures more complex, non-linear sensor patterns
2	Wider hidden layers (7 → 14 → 32 units)	Models 1–6	Increases model capacity to learn richer representations
3	Dropout regularization (rate = 0.5)	Models 2, 3, 5, 6	Reduces overfitting by randomly deactivating neurons
4	Class weights (inverse frequency)	Models 3, 6	Up-weights the minority failure class during training
5	SGD optimizer	Models 0–3, 6	Conservative updates; robust with proper tuning
6	Adam optimizer	Models 4, 5	Adaptive learning rates; faster convergence on imbalanced data
7	Architecture search (comparing all variants)	All	Identifies the best depth/width/regularization combination

	Model 0	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Accuracy	0.988688	0.991313	0.989000	0.990313	0.993563	0.992500	0.990625
Recall	0.908685	0.929684	0.908321	0.941344	0.944654	0.937732	0.950519
Precision	0.981335	0.986299	0.985233	0.964564	0.993633	0.990039	0.959293
F1 Score	0.941668	0.956017	0.943049	0.952616	0.967701	0.962222	0.954858

	Model 0	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Accuracy	0.988000	0.988250	0.988250	0.987750	0.991250	0.991000	0.989000
Recall	0.904611	0.908983	0.902624	0.934158	0.929651	0.929518	0.934819
Precision	0.978365	0.976234	0.983438	0.947464	0.985670	0.983170	0.958211
F1 Score	0.938015	0.939726	0.938881	0.940696	0.955726	0.954564	0.946167

Criterion	Model 4	Runner-up (Model 1)	Notes
Val F1 (macro)	0.961	0.930	Highest F1 balances precision and recall
Val Recall (macro)	0.957	0.896	Critical — missed failures are costly
Val Accuracy	0.973	0.951	Consistent with F1 ranking
Overfitting	Minimal gap	Moderate gap	Train ≈ Val metrics indicate good generalization
Architecture	14 → 7 (simpler)	14 → 7	Same architecture, Adam converges faster