📈 Sales Forecasting — Time Series Prediction

🎯 Objective

Analyze historical sales data and generate future demand forecasts using time-series techniques.

🧰 Tech Stack

Python, pandas, matplotlib, statsmodels, scikit-learn

⚙️ Approach

• Converted dates to datetime and ordered chronologically
• Split data into train (80%) and test (20%)
• Applied Exponential Smoothing
• Evaluated using MAE and MAPE
• Exported predictions to CSV

📊 Results

• MAE: 241,823.39
• MAPE: Infinite (caused by zero actual values in some periods)

The model follows the overall movement but struggles during high volatility.

📁 Files

• task16_forecasting.ipynb → implementation
• forecast_output.csv → predictions
• forecast_report.txt → detailed findings

🚀 Future Work

Improve accuracy by handling zero-demand cases and incorporating seasonality or external drivers.

👩‍💻 Author

Soumya Pandey