2.-LSTM-Model.md

+
+**Outline of model training**
+    
+**default**
+4 different locations
+
+epochs=200
+
+batch_size=64
+
+callbacks=[checkpoint, early_stopping (patience=5), reduce_lr(factor=0.1, patience=5, min_lr=1e-6)]
+
+learning_rate=0.001
+
+LSTM layer with 128 units
+
+LSTM output layer
+
+**testing**
+
+LSTM units 128, 64, 32
+
+Dense units 128, 64, 32
+
+bidir units 128, 64
+
+batch_size 128, 64, 32
+
+2 lstm layers
+
+learning rate
+
+dropout layer
+
+patience
+
+combinations of the parameters with better results
+
+```
+def train_model(model, X_train, y_train, X_test, y_test):
+    epochs=200
+    batch_size=64
+
+    early_stopping = EarlyStopping(monitor='val_loss', patience=5, restore_best_weights=True)
+
+    # Learning rate scheduler callback
+    reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, min_lr=1e-6)
+
+
+    checkpoint_path = "checkpoints/model_epoch_{epoch:02d}_val_loss_{val_loss:.2f}.weights.h5"
+    checkpoint = ModelCheckpoint(
+        filepath=checkpoint_path, save_weights_only=True,
+        monitor='val_loss', mode='min', save_best_only=True, verbose=1
+    )
+
+    history = model.fit(X_train, y_train, epochs=epochs, batch_size=batch_size,
+           validation_data=(X_test, y_test), callbacks=[checkpoint, early_stopping, reduce_lr])
+    return model, history
+
+
+model1 = Sequential([
+    Input(shape=(sequence_length, feature_count)),
+    LSTM(128, return_sequences=True),
+    LSTM(sequence_length) # Ensure the output shape matches y
+])
+
+optimizer = Adam(learning_rate=0.001, clipvalue=1.0)
+```
+
+**testing of different Parameters and Layers**
+