build_lstm.py

import numpy as np
from sklearn.model_selection import train_test_split
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense

def preprocess_data(X_raw):
    """
    Normalize the raw data for LSTM processing.
    
    Parameters:
        X_raw (array): The raw feature set.
        
    Returns:
        array: Normalized feature set.
    """
    # Min-max scaling for normalization
    X_min = np.min(X_raw, axis=0)
    X_max = np.max(X_raw, axis=0)
    X_normalized = (X_raw - X_min) / (X_max - X_min)
    
    return X_normalized

def build_lstm(input_shape):
    """
    Build a Long Short-Term Memory (LSTM) model for processing EHR data.
    
    Parameters:
        input_shape (tuple): The shape of the input sequences.
        
    Returns:
        Sequential: A simple LSTM model.
    """
    model = Sequential()
    model.add(LSTM(50, input_shape=input_shape))
    model.add(Dense(1, activation='sigmoid'))
    
    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
    return model

def train_and_evaluate_model(model, X_train, y_train, X_val, y_val):
    """
    Train and evaluate the LSTM model.
    
    Parameters:
        model (Sequential): The LSTM model.
        X_train (array): Training feature set.
        y_train (array): Training labels.
        X_val (array): Validation feature set.
        y_val (array): Validation labels.
        
    Returns:
        dict: Training history.
    """
    history = model.fit(X_train, y_train, epochs=10, validation_data=(X_val, y_val))
    return history.history

# Generate synthetic data for demonstration
X_raw = np.random.rand(100, 10, 5)  # 100 samples, 10 timesteps, 5 features
y_raw = np.random.randint(2, size=100)  # 100 labels (0 or 1)

# Preprocess the data
X_processed = preprocess_data(X_raw)

# Split the data into training and validation sets
X_train, X_val, y_train, y_val = train_test_split(X_processed, y_raw, test_size=0.2, random_state=42)

# Build the LSTM model
input_shape = (X_train.shape[1], X_train.shape[2])
model = build_lstm(input_shape)

# Train and evaluate the model
history = train_and_evaluate_model(model, X_train, y_train, X_val, y_val)

# Output the training history
print("Training History:", history)