main.py

import os
import pandas as pd
import joblib

from src.data_preprocessing import load_data, preprocess
from src.feature_engineering import create_features
from src.forecasting import train_model, predict
from src.inventory import calculate_safety_stock, reorder_point
from src.visualization import plot_sales

# -----------------------------
# Step 1: Fix file path
# -----------------------------
base_dir = os.path.dirname(os.path.abspath(__file__))
file_path = os.path.join(base_dir, "data", "raw", "sales.csv")

# -----------------------------
# Step 2: Load & preprocess data
# -----------------------------
df = load_data(file_path)
df = preprocess(df)

# Keep original Product_ID
df['Product_ID_Original'] = df['Product_ID']

# Encode Product_ID for model
df['Product_ID_Encoded'] = df['Product_ID'].astype('category').cat.codes

# Save processed data
processed_path = os.path.join(base_dir, "data", "processed", "cleaned_sales.csv")
df.to_csv(processed_path, index=False)

# -----------------------------
# Step 3: Feature engineering
# -----------------------------
df = create_features(df)

# -----------------------------
# Step 4: Train model
# -----------------------------
X = df[['day', 'month', 'weekday', 'Product_ID_Encoded']]
y = df['Sales']

model = train_model(X, y)

# Save model
model_path = os.path.join(base_dir, "models", "model.pkl")
joblib.dump(model, model_path)

# -----------------------------
# Step 5: Predictions
# -----------------------------
predictions = predict(model, X)

# Save forecast output
forecast_df = pd.DataFrame({
    "Date": df["Date"],
    "Product_ID": df["Product_ID_Original"],
    "Actual": y,
    "Predicted": predictions
})

forecast_path = os.path.join(base_dir, "outputs", "forecasts.csv")
forecast_df.to_csv(forecast_path, index=False)

# -----------------------------
# Step 6: Inventory calculations (USING PREDICTIONS)
# -----------------------------
avg_demand = predictions.mean()
std_demand = predictions.std()

lead_time = 5

safety = calculate_safety_stock(std_demand, lead_time)
reorder = reorder_point(avg_demand, lead_time, safety)

# Save inventory output
inventory_df = pd.DataFrame({
    "Average Demand": [avg_demand],
    "Demand Std Dev": [std_demand],
    "Safety Stock": [safety],
    "Reorder Point": [reorder]
})

inventory_path = os.path.join(base_dir, "outputs", "inventory_plan.csv")
inventory_df.to_csv(inventory_path, index=False)

# -----------------------------
# Step 7: Visualization
# -----------------------------
plot_sales(df)

# -----------------------------
# Step 8: Print results
# -----------------------------
print("Average Demand:", avg_demand)
print("Demand Std Dev:", std_demand)
print("Safety Stock:", safety)
print("Reorder Point:", reorder)

# -----------------------------
# Step 9: Future Forecasting (FIXED)
# -----------------------------

future_days = 7
last_date = df['Date'].max()

future_data = []

products = df['Product_ID_Original'].unique()

for product in products:
    for i in range(1, future_days + 1):
        future_date = last_date + pd.Timedelta(days=i)

        future_data.append({
            "Date": future_date,
            "Product_ID_Original": product
        })

future_df = pd.DataFrame(future_data)

# ✅ IMPORTANT: encode product SAME as training
future_df['Product_ID_Encoded'] = future_df['Product_ID_Original'].astype('category').cat.codes

# Feature engineering
future_df['day'] = future_df['Date'].dt.day
future_df['month'] = future_df['Date'].dt.month
future_df['weekday'] = future_df['Date'].dt.weekday

# Prediction input MUST match training features
X_future = future_df[['day', 'month', 'weekday', 'Product_ID_Encoded']]

future_df['Predicted_Sales'] = model.predict(X_future)

# Save output
future_path = os.path.join(base_dir, "outputs", "future_forecast.csv")
future_df.to_csv(future_path, index=False)

print("Future Forecast Saved!")