# Let us get Python ready for time series forecasting.
import warnings; warnings.filterwarnings("ignore")
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

# Data setup: let us load a real retail sales dataset for forecasting.
url = "https://raw.githubusercontent.com/jbrownlee/Datasets/master/shampoo.csv"
df = pd.read_csv(url)
print("Shape:", df.shape)
print("Preview:")
print(df.head())

# Let us plot the sales numbers to see the patterns.
plt.figure(figsize=(8,4))
plt.plot(df["Sales"])
plt.title("Monthly Shampoo Sales")
plt.xlabel("Month")
plt.ylabel("Sales")
plt.show()

# Fix missing values and clean up column names if needed.
df = df.rename(columns={"Month":"month","Sales":"sales"})
df["sales"] = pd.to_numeric(df["sales"], errors="coerce")
df = df.dropna().reset_index(drop=True)
print(df.head())

# Let us create sequences for Seq2Seq forecasting.
def create_sequences(data, input_size=3, output_size=2):
    X, y = [], []
    for i in range(len(data) - input_size - output_size + 1):
        X.append(data[i:i+input_size])
        y.append(data[i+input_size:i+input_size+output_size])
    return np.array(X), np.array(y)
sales_data = df["sales"].values
X, y = create_sequences(sales_data, input_size=3, output_size=2)
print("Input shape:", X.shape, "Output shape:", y.shape)

# Look at some sample input-output sequence pairs.
for i in range(3):
    print("Input:", X[i], "-> Output:", y[i])
    

# Let us split the data into train and test sets for fair testing.
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)
print("Training samples:", X_train.shape[0])
print("Testing samples:", X_test.shape[0])

# Build a super simple sequence-to-sequence model: Linear Regression.
from sklearn.linear_model import LinearRegression
model = LinearRegression()
X_train_flat = X_train.reshape(X_train.shape[0], -1)
y_train_flat = y_train.reshape(y_train.shape[0], -1)
model.fit(X_train_flat, y_train_flat)

# Make predictions on our test set and see results.
X_test_flat = X_test.reshape(X_test.shape[0], -1)
y_pred = model.predict(X_test_flat)
for i in range(3):
    print("Predicted:", np.round(y_pred[i],1), "| True:", y_test[i])
    

# Plot predictions versus true values for a better picture.
plt.figure(figsize=(8,4))
plt.plot(np.arange(len(y_test)), y_test[:,0], label="True First Month")
plt.plot(np.arange(len(y_test)), y_pred[:,0], label="Predicted First Month")
plt.title("Forecast: True vs Predicted (First Month Ahead)")
plt.legend()
plt.show()

# Calculate a simple accuracy score for our predictions.
from sklearn.metrics import mean_squared_error
mse = mean_squared_error(y_test, y_pred)
print("Mean Squared Error:", round(mse,2))

# Practice: Try your own sequence. Enter some test sales values.
user_input = input("Type 3 sales values, separated by commas: ")
vals = [float(x.strip()) for x in user_input.split(",")]
vals_arr = np.array(vals).reshape(1,-1)
pred = model.predict(vals_arr)
print("Your input:", vals)
print("Forecasted next two months:", np.round(pred[0],1))

# Mini-project: Forecast next month based on the past year.
past_12 = df["sales"].values[-12:]
X_project = past_12[-3:].reshape(1, -1)
future_pred = model.predict(X_project)
print("Recent three months:", past_12[-3:])
print("Predicted next two months:", np.round(future_pred[0],1))

# Challenge: Change to a different input size and re-train.
X2, y2 = create_sequences(sales_data, input_size=6, output_size=2)
X2_train, X2_test, y2_train, y2_test = train_test_split(X2, y2, test_size=0.2, random_state=42)
model2 = LinearRegression()
model2.fit(X2_train.reshape(X2_train.shape[0], -1), y2_train.reshape(y2_train.shape[0], -1))
score2 = model2.score(X2_test.reshape(X2_test.shape[0], -1), y2_test)
print("Model accuracy with input_size=6:", round(score2,2))

# Extra tips: Try a more advanced model if you like challenges.
from sklearn.ensemble import RandomForestRegressor
rf = RandomForestRegressor(random_state=42)
rf.fit(X_train_flat, y_train_flat)
rf_pred = rf.predict(X_test_flat)
rf_mse = mean_squared_error(y_test, rf_pred)
print("Random Forest Mean Squared Error:", round(rf_mse,2))

# Trouble? Here is a quick check for data shapes.
print("X_train shape:", X_train.shape)
print("y_train shape:", y_train.shape)
print("X_test shape:", X_test.shape)
print("y_test shape:", y_test.shape)