Demand Forecasting (Time-Series)

End-to-end demand forecasting with Python using synthetic time-series sales data. Includes data generation, cleaning, ARIMA/SARIMA model selection by AIC, evaluation with RMSE and MAPE, and 90-day forecasts with confidence intervals. Reproducible scripts and visualizations for portfolio showcase.

Forecast daily sales with ARIMA/SARIMA on synthetic data. Includes generation, cleaning, train/validation split, AIC-based model search, evaluation (RMSE/MAPE), and 90-day forecast with confidence intervals. Production-ready scripts and charts for portfolio showcase.

---

Features

• Synthetic daily sales generator (trend + weekly + annual seasonality + noise)
• Train/validation split
• ARIMA/SARIMA model search by AIC (weekly seasonality)
• Metrics: RMSE, MAPE
• 90-day forecast with confidence intervals
• Plots: history vs. forecast, residual diagnostics
• Deterministic seeding for reproducibility

---

Project Structure

demand-forecasting/
├─ README.md
├─ LICENSE
├─ requirements.txt
├─ data/
│  └─ generate_timeseries.py
├─ src/
│  ├─ forecast_arima.py
│  └─ metrics.py
└─ outputs/
   └─ figures & reports

---

Setup

python -m venv .venv
# Windows:
.venv\Scripts\activate
# macOS/Linux:
source .venv/bin/activate
pip install -r requirements.txt

---

Generate Synthetic Data

python data/generate_timeseries.py --start 2023-01-01 --end 2024-12-31 --seed 42 --out data/daily_sales.csv

---

Run Forecast

python src/forecast_arima.py --input data/daily_sales.csv --horizon 90 --val_days 60 --outdir outputs

Outputs

• outputs/metrics.json – RMSE & MAPE (validation)
• outputs/forecast.csv – point forecast + confidence intervals
• outputs/fig_history_forecast.png
• outputs/fig_residuals.png

---

Sample Results

Forecast vs History

Residual Diagnostics

Key Metrics

Metric	Value
RMSE	2.11
MAPE	2.77%
ARIMA Order	(2,1,2)
Seasonal Order	(0,1,1,7)
AIC	2836.7

---

Data Schema

column	description
date	daily timestamp
sales	units sold (int)