❮ Previous

Model Implementation

  • This lesson explains how machine learning and deep learning models are implemented and used in real-world systems.

  • Model Implementation in Deep Learning (ANN Example)

    In this module, we will cover the complete workflow:

    Building ANN Model
    Model Compilation
    Model Training
    Model Evaluation
    Saving & Loading Model

    We’ll use a simple Binary Classification example.


    Building ANN Model

    We will create a simple Artificial Neural Network using TensorFlow / Keras.

    Example: Binary Classification

Building a Simple ANN for Binary Classification using TensorFlow Keras

This Python example demonstrates how to build a feedforward Artificial Neural Network (ANN) for binary classification using TensorFlow Keras. The model consists of an input layer for 20 features, two hidden Dense layers with ReLU activation, and an output layer with Sigmoid activation for predicting binary outcomes. 

import tensorflow as tf
from tensorflow.keras import layers, models

# Build ANN model
model = models.Sequential([
    layers.Dense(64, activation='relu', input_shape=(20,)),  # 20 input features
    layers.Dense(32, activation='relu'),
    layers.Dense(1, activation='sigmoid')  # Binary output
])

  • Explanation:

    • Dense(64) → Hidden layer

    • ReLU → Activation function

    • Dense(1) → Output layer

    • Sigmoid → For binary classification


    Model Compilation

    Compilation defines:

    Optimizer
    Loss Function
    Evaluation Metric

Compiling a Binary Classification ANN in TensorFlow Keras

This Python example demonstrates how to compile a binary classification neural network using TensorFlow Keras. The model uses the Adam optimizer, binary_crossentropy as the loss function for binary outcomes, and accuracy as the evaluation metric to track performance during training. 

model.compile(
    optimizer='adam',
    loss='binary_crossentropy',
    metrics=['accuracy']
)

  • Why These?

    • Adam → Fast & adaptive optimizer

    • Binary Crossentropy → Binary classification loss

    • Accuracy → Evaluation metric


    Model Training

    We train model using .fit().

Training a Binary Classification ANN in TensorFlow Keras

This Python example demonstrates how to train a binary classification neural network using TensorFlow Keras. The model is trained on the training dataset for multiple epochs with a specified batch size, while using validation data to monitor performance. The training history stores metrics such as loss and accuracy for both training and validation sets. 

history = model.fit(
    X_train, y_train,
    epochs=20,
    batch_size=32,
    validation_data=(X_val, y_val)
)

  • Parameters:

    • epochs → Number of full passes over dataset

    • batch_size → Number of samples per update

    • validation_data → To monitor overfitting

    Sample Output:

    Epoch 1/20

    loss: 0.65 - accuracy: 0.70 - val_loss: 0.60 - val_accuracy: 0.75


    Epoch 2/20

    loss: 0.50 - accuracy: 0.82 - val_loss: 0.45 - val_accuracy: 0.85



    Model Evaluation

    Evaluate performance on test data.

    test_loss, test_accuracy = model.evaluate(X_test, y_test)

    print("Test Accuracy:", test_accuracy)



    Other Evaluation Methods:


    predictions = model.predict(X_test)


    For classification:

    • Accuracy

    • Precision

    • Recall

    • F1-score

    • Confusion Matrix


    Saving & Loading Model

    Very important for real projects.

    Save Model

    model.save("ann_model.h5")

    OR newer format:

    model.save("ann_model.keras")

    Load Model

Loading a Saved Keras Model and Making Predictions in Python

This Python example demonstrates how to load a previously saved TensorFlow Keras model from an .h5 file and use it to make predictions. After loading the model with load_model(), you can call predict() on new input data (X_test) to generate predictions without retraining the model. 

from tensorflow.keras.models import load_model

loaded_model = load_model("ann_model.h5")

Now you can use:
loaded_model.predict(X_test)
❮ Previous