Version 2 con funciones incluidas

keras_model_functions.py

@@ -57,3 +57,47 @@ def train_model(model, X_train, y_train, X_val, y_val, epochs=200, batch_size=32
     )
 
     return history
+
+def add_brownian_noise(X):
+    brownian_noise = tf.random.normal(tf.shape(X), mean=0.0, stddev=0.1, dtype=tf.float32)
+    return X + brownian_noise
+
+def evaluate_model(model, X_test, y_test):
+    result = model.evaluate(X_test, y_test, verbose=0)
+    print("Pérdida en el conjunto de prueba:", result)
+
+def print_model_info(model):
+    print("\nCaracterísticas del modelo:")
+    print("Número de capas:", len(model.layers))
+    model.summary()
+
+def calculate_accuracy(y_true, y_pred):
+    tasa_acierto = mean_absolute_error(y_test, y_pred)
+    print("Tasa de acierto (MAE):", tasa_acierto)
+
+def plot_predictions(y_test, y_pred, df):
+    plt.figure(figsize=(10, 6))
+    plt.plot(y_test, label='Valor Real')
+    plt.plot(y_pred, label='Predicción', alpha=0.7)
+    plt.title("Comparación de Predicciones vs. Valoreseales")
+    plt.xlabel("Índice")
+    plt.ylabel("Precio Escalado")
+    plt.legend()
+    plt.show()
+
+# Hacer predicciones
+train_predict = model.predict(X_train)
+test_predict = model.predict(X_test)
+
+# Invertir las predicciones a la escala original
+train_predict = scaler.inverse_transform(np.hstack([train_predict, np.zeros((train_predict.shape[0], X.shape[2]-1))]))[:, 0]
+test_predict = scaler.inverse_transform(np.hstack([test_predict, np.zeros((test_predict.shape[0], X.shape[2]-1))]))[:, 0]
+y_train_inv = scaler.inverse_transform(np.hstack([y_train.reshape(-1, 1), np.zeros((y_train.shape[0], X.shape[2]-1))]))[:, 0]
+y_test_inv = scaler.inverse_transform(np.hstack([y_test.reshape(-1, 1), np.zeros((y_test.shape[0], X.shape[2]-1))]))[:, 0]
+
+# Calcular el error cuadrático medio (RMSE)
+train_rmse = np.sqrt(np.mean((train_predict - y_train_inv)**2))
+test_rmse = np.sqrt(np.mean((test_predict - y_test_inv)**2))
+
+print(f"RMSE en entrenamiento: {train_rmse}")
+print(f"RMSE en prueba: {test_rmse}")