import warnings; warnings.filterwarnings("ignore")
# Let us load some basic helper modules
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

# Download and load the airline passengers dataset
url = "https://raw.githubusercontent.com/jbrownlee/Datasets/master/airline-passengers.csv"
df = pd.read_csv(url, parse_dates=["Month"])
print("Data shape:", df.shape)
df.head()

# Show the time series as a line chart
plt.figure(figsize=(8,4))
plt.plot(df["Month"], df["Passengers"])
plt.title("Monthly Airline Passengers")
plt.xlabel("Month")
plt.ylabel("Passengers")
plt.show()

# Let us prepare the data as numbers for machine learning
series = df["Passengers"].values.astype(np.float32)
print(series[:10])

# Normalize (scale) the data to 0-1 range
from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler()
series_scaled = scaler.fit_transform(series.reshape(-1,1)).flatten()
print(series_scaled[:10])

# Split into training and test sets
split = int(len(series_scaled) * 0.8)
train, test = series_scaled[:split], series_scaled[split:]
print(f"Train size: {len(train)}  Test size: {len(test)}")

# Create features and labels for supervised learning
def make_windows(series, window_size):
    windows = []
    labels = []
    for i in range(len(series) - window_size):
        windows.append(series[i:i+window_size])
        labels.append(series[i+window_size])
    return np.array(windows), np.array(labels)

window_size = 12  # 12 months = 1 year
X_train, y_train = make_windows(train, window_size)
X_test, y_test = make_windows(test, window_size)
print("Train window shape:", X_train.shape)

# Import TensorFlow and Keras layers
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers

# Prepare training data shape for CNN: (samples, steps, features)
X_train_cnn = X_train[..., np.newaxis]
X_test_cnn = X_test[..., np.newaxis]
print("CNN input shape:", X_train_cnn.shape)

# Build a simple 1D CNN
model = keras.Sequential([
    layers.Conv1D(32, kernel_size=3, activation="relu", input_shape=(window_size, 1)),
    layers.MaxPooling1D(pool_size=2),
    layers.Flatten(),
    layers.Dense(32, activation="relu"),
    layers.Dense(1)
])
model.compile(optimizer="adam", loss="mse")
model.summary()

# Train the CNN model
history = model.fit(X_train_cnn, y_train, epochs=40, batch_size=8, validation_data=(X_test_cnn, y_test), verbose=0)
print("Training complete.")

# Plot the training history
plt.plot(history.history["loss"], label="Train Loss")
plt.plot(history.history["val_loss"], label="Val Loss")
plt.title("Training Loss over Epochs")
plt.xlabel("Epoch")
plt.ylabel("Loss (MSE)")
plt.legend()
plt.show()

# Make predictions using the trained model
preds = model.predict(X_test_cnn).flatten()
# Reverse scaling to get back to the real passenger numbers
y_test_real = scaler.inverse_transform(y_test.reshape(-1,1)).flatten()
preds_real = scaler.inverse_transform(preds.reshape(-1,1)).flatten()

# Plot predictions vs actual values
plt.figure(figsize=(8,4))
plt.plot(y_test_real, label="Actual")
plt.plot(preds_real, label="Predicted")
plt.title("CNN Predictions vs Actual Passengers")
plt.xlabel("Time Step")
plt.ylabel("Passengers")
plt.legend()
plt.show()

# Calculate how good our model is (Mean Squared Error)
from sklearn.metrics import mean_squared_error
mse = mean_squared_error(y_test_real, preds_real)
print(f"Mean Squared Error on Test Data: {mse:.2f}")

# Best practices: Avoid overfitting with Dropout layers
model2 = keras.Sequential([
    layers.Conv1D(32, kernel_size=3, activation="relu", input_shape=(window_size, 1)),
    layers.MaxPooling1D(pool_size=2),
    layers.Dropout(0.2),
    layers.Flatten(),
    layers.Dense(32, activation="relu"),
    layers.Dropout(0.2),
    layers.Dense(1)
])
model2.compile(optimizer="adam", loss="mse")
model2.summary()

# Troubleshooting: If you get errors about input shapes, print your data shapes.
print("X_train_cnn shape:", X_train_cnn.shape)
print("y_train shape:", y_train.shape)

# Tip: Use input() to ask the user for the window size.
user_window = input("Enter the number of months to use for each window (example: 12): ")
user_window = int(user_window)
X_train2, y_train2 = make_windows(train, user_window)
print(f"New shape: {X_train2.shape}")