Demand Forecasting (Python)

Project Overview

This project forecasts monthly product demand using simple time-series methods. The goal is to demonstrate practical forecasting techniques for inventory and supply chain planning.

Two models are implemented and compared:

• Moving Average
• Simple Exponential Smoothing

The best-performing model is then used to forecast future demand.

---

Dataset

The original dataset contains daily demand data for multiple SKUs.
The data was aggregated into monthly demand for a single product: SKU-004.

---

Methodology

1. Train–Test Split

• First 9 months: training data
• Last 3 months: test data

This allows evaluation of model accuracy on unseen data.

---

2. Forecasting Models

Moving Average

A 3-month moving average is used:

Forecast = average of last 3 months

Exponential Smoothing

Simple exponential smoothing with:

α = 0.3

This method gives more weight to recent observations.

---

3. Model Evaluation

Two error metrics are used:

• MAE (Mean Absolute Error)
• MAPE (Mean Absolute Percentage Error)

Model	MAE	MAPE
Moving Average	24.10	3.83%
Exponential Smoothing	20.98	3.32%

The exponential smoothing model performed better.

---

Forecast Results

Using the best model, the forecasted demand is:

Month	Forecast
Jan 2026	612 units
Feb 2026	612 units
Mar 2026	612 units

---

Project Structure

demand-forecasting-python/
│
├── data/
│   └── inventory_sales_daily.xlsx
│
├── outputs/
│   ├── monthly_series_SKU-004.xlsx
│   ├── monthly_series_SKU-004.png
│   ├── model_comparison_SKU-004.xlsx
│   ├── forecast_test_comparison_SKU-004.png
│   └── future_forecast_SKU-004_*.xlsx
│
├── demand_forecasting_report.pdf
├── demand_forecasting.py
├── requirements.txt
└── README.md

---

Tools Used

• Python
• pandas
• numpy
• matplotlib
• Excel

---

How to Run

1. Install dependencies:

pip install -r requirements.txt

2. Run the script:

python demand_forecasting.py

Outputs will be saved in:

outputs/

---

Author

Parnia Riazat