What Harsh demonstrated

Built an end-to-end demand forecasting workflow on 9,648 historical transactions: resampled sales to monthly frequency, decomposed trend and seasonality, compared moving average, linear regression, and SARIMA models, and evaluated them with MAE, RMSE, and MAPE on a true future holdout. The final comparison showed SARIMA as the strongest forecast, supported by clear visual and metric-based analysis.

• Leakage-Free Time-Series Splitting
• Comparative Forecast Evaluation
• Seasonality and Trend Detection
• Business-Oriented Insight Extraction