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.\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n**Environment Setup:**\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`bash\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\npip install pandas numpy scikit-learn prophet matplotlib seaborn joblib fastapi uvicorn\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n**Part 1 — Data Preparation:**\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`python\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nimport pandas as pd\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nimport numpy as np\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nfrom sklearn.preprocessing import LabelEncoder\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Load sales data\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf = pd.read_csv('sales.csv', parse_dates=['date'])\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Feature engineering for time series\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['year'] = df['date'].dt.year\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['month'] = df['date'].dt.month\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['week'] = df['date'].dt.isocalendar().week\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['day_of_week'] = df['date'].dt.dayofweek\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['quarter'] = df['date'].dt.quarter\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Lag variables (demand from previous periods)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['demand_lag_1'] = df.groupby('sku')['demand'].shift(1)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['demand_lag_4'] = df.groupby('sku')['demand'].shift(4)  # 4 weeks ago\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['rolling_avg_4'] = df.groupby('sku')['demand'].transform(\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    lambda x: x.shift(1).rolling(4).mean()\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Holiday indicators (adjust for your market)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nholidays = ['2025-01-01', '2025-04-18', '2025-09-07', '2025-12-25']\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['is_holiday'] = df['date'].dt.date.astype(str).isin(holidays).astype(int)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Encode categorical variables\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nle = LabelEncoder()\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['sku_encoded'] = le.fit_transform(df['sku'])\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndf['category_encoded'] = le.fit_transform(df['category'])\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n**Part 2 — Regression Model (scikit-learn):**\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`python\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nfrom sklearn.ensemble import GradientBoostingRegressor\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nfrom sklearn.model_selection import TimeSeriesSplit\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nfrom sklearn.metrics import mean_absolute_percentage_error, mean_squared_error\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Features and target\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nfeatures = ['sku_encoded', 'month', 'week', 'day_of_week', 'quarter',\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n            'demand_lag_1', 'demand_lag_4', 'rolling_avg_4', 'is_holiday',\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n            'unit_price', 'discount_pct']\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nX = df[features].dropna()\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ny = df.loc[X.index, 'demand']\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Cross-validation respecting temporal order\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ntscv = TimeSeriesSplit(n_splits=5)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nerrors = []\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nfor train_idx, val_idx in tscv.split(X):\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    X_train, X_val = X.iloc[train_idx], X.iloc[val_idx]\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    y_train, y_val = y.iloc[train_idx], y.iloc[val_idx]\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    model = GradientBoostingRegressor(\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n        n_estimators=300, max_depth=5, learning_rate=0.05, random_state=42\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    )\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    model.fit(X_train, y_train)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    predictions = model.predict(X_val)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    errors.append(mean_absolute_percentage_error(y_val, predictions))\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nprint(f'Average MAPE: {np.mean(errors):.2%}')  # Target: < 15% for perishables\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Train final model with all data\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nfinal_model = GradientBoostingRegressor(...).fit(X, y)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n# Feature importance for explainability\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nimportance = pd.Series(final_model.feature_importances_, index=features)\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nprint(importance.sort_values(ascending=False))\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n**Part 3 — Facebook Prophet for High-Volume SKUs:**\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\`python\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\nfrom prophet import Prophet\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ndef forecast_with_prophet(df_sku, periods=8):\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\n    \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\