Predictive Analysis for Demand Forecasting and Inventory Optimization

This project focuses on analyzing historical product demand data, building demand forecasting workflows, and applying inventory optimization techniques to support better supply chain decisions.

The work is presented through Jupyter notebooks and combines data cleaning, exploratory analysis, machine learning, and linear programming based optimization.

Overview

The main goal of this project is to use historical order demand data to:

• understand demand patterns across products and warehouses
• prepare data for predictive modeling
• compare forecasting approaches
• optimize inventory allocation and replenishment decisions
• evaluate scenario and sensitivity changes in demand

Key Features

• end-to-end demand data preprocessing
• exploratory data analysis with trend and distribution plots
• feature engineering for predictive modeling
• model comparison using Linear Regression, Decision Tree, and Random Forest
• inventory optimization using PuLP
• sensitivity analysis for changing demand conditions

Project Files

• PredictiveAnaOptimized.ipynb
  Refined and validated notebook with forecasting and optimization workflow.

• PredictiveAnalysis.ipynb
  Original notebook covering exploratory analysis, model training, and optimization experiments.

• Historical Product Demand.csv
  Raw dataset used for the analysis.

• data_processed.csv
  Processed dataset derived from the source file.

Tools and Libraries

• Python
• Jupyter Notebook
• pandas
• numpy
• matplotlib
• seaborn
• scikit-learn
• statsmodels
• PuLP

Setup

Clone the repository and install dependencies:

pip install -r requirements.txt

Launch Jupyter Notebook:

jupyter notebook

Open either notebook from the project folder and run the cells in order.

Results Summary

This project demonstrates how predictive analytics can be used to connect historical demand patterns with inventory planning decisions. The optimized notebook successfully runs through forecasting, optimization, and comparative evaluation workflows in the validated environment.

Validation

• PredictiveAnaOptimized.ipynb was executed successfully during project verification.
• PredictiveAnalysis.ipynb was reviewed and fixed so its workflow completes correctly.

Future Improvements

• convert notebook logic into modular Python scripts
• add clearer business assumptions for the optimization model
• create a lightweight version without large dataset files
• build a simple dashboard for results visualization

License

This project is licensed under the MIT License.