Predictive Analysis with Python: Regression and Demand Forecasting

Create a complete demand forecasting pipeline in Python for [COMPANY NAME], with [NUMBER] SKUs and [PERIOD] months of sales history.

**Environment Setup:**
```bash
pip install pandas numpy scikit-learn prophet matplotlib seaborn joblib fastapi uvicorn
```

**Part 1 — Data Preparation:**
```python
import pandas as pd
import numpy as np
from sklearn.preprocessing import LabelEncoder

# Load sales data
df = pd.read_csv('sales.csv', parse_dates=['date'])

# Feature engineering for time series
df['year'] = df['date'].dt.year
df['month'] = df['date'].dt.month
df['week'] = df['date'].dt.isocalendar().week
df['day_of_week'] = df['date'].dt.dayofweek
df['quarter'] = df['date'].dt.quarter

# Lag variables (demand from previous periods)
df['demand_lag_1'] = df.groupby('sku')['demand'].shift(1)
df['demand_lag_4'] = df.groupby('sku')['demand'].shift(4)  # 4 weeks ago
df['rolling_avg_4'] = df.groupby('sku')['demand'].transform(
    lambda x: x.shift(1).rolling(4).mean()
)

# Holiday indicators (adjust for your market)
holidays = ['2025-01-01', '2025-04-18', '2025-09-07', '2025-12-25']
df['is_holiday'] = df['date'].dt.date.astype(str).isin(holidays).astype(int)

# Encode categorical variables
le = LabelEncoder()
df['sku_encoded'] = le.fit_transform(df['sku'])
df['category_encoded'] = le.fit_transform(df['category'])
```

**Part 2 — Regression Model (scikit-learn):**
```python
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.model_selection import TimeSeriesSplit
from sklearn.metrics import mean_absolute_percentage_error, mean_squared_error

# Features and target
features = ['sku_encoded', 'month', 'week', 'day_of_week', 'quarter',
            'demand_lag_1', 'demand_lag_4', 'rolling_avg_4', 'is_holiday',
            'unit_price', 'discount_pct']
X = df[features].dropna()
y = df.loc[X.index, 'demand']

# Cross-validation respecting temporal order
tscv = TimeSeriesSplit(n_splits=5)

errors = []
for train_idx, val_idx in tscv.split(X):
    X_train, X_val = X.iloc[train_idx], X.iloc[val_idx]
    y_train, y_val = y.iloc[train_idx], y.iloc[val_idx]
    
    model = GradientBoostingRegressor(
        n_estimators=300, max_depth=5, learning_rate=0.05, random_state=42
    )
    model.fit(X_train, y_train)
    predictions = model.predict(X_val)
    errors.append(mean_absolute_percentage_error(y_val, predictions))

print(f'Average MAPE: {np.mean(errors):.2%}')  # Target: < 15% for perishables

# Train final model with all data
final_model = GradientBoostingRegressor(...).fit(X, y)

# Feature importance for explainability
importance = pd.Series(final_model.feature_importances_, index=features)
print(importance.sort_values(ascending=False))
```

**Part 3 — Facebook Prophet for High-Volume SKUs:**
```python
from prophet import Prophet

def forecast_with_prophet(df_sku, periods=8):