Retail Prediction

AI-based demand forecasting application for retail inventory optimization using statistical and Machine Learning models.

Project Overview

This project was developed as part of my Master's dissertation and aims to improve inventory planning by forecasting future product demand using historical retail sales data.

Instead of relying only on average daily sales, the application evaluates multiple forecasting algorithms and compares their performance to support inventory replenishment decisions.

The project includes the complete data science workflow:

Data extraction

Data preprocessing

Exploratory Data Analysis (EDA)

Time Series Analysis

Forecasting models

Model evaluation

Interactive dashboard

Technologies

• Python
• Pandas
• NumPy
• Scikit-Learn
• Statsmodels
• TensorFlow / Keras
• XGBoost
• Prophet
• Matplotlib
• Streamlit

Pre - processing

Exploratory Analysis

Sales Evolution Pearson Correlation - Heatmap

STL Decomposition ( observed, trend, season, resid )

Observed, Trend, Seasonal and Residual components were extracted to better understand the time series behaviour.

Train Test Split

Forecasting Models

The project compares both statistical and Machine Learning approaches.

Model Evaluation

Models were compared using:

MAE RMSE R² Score

Performance comparison was used to determine the most suitable forecasting model for retail demand prediction.

xgboost

Dashboard

The application includes an interactive Streamlit dashboard where users can:

explore historical sales visualize forecasts compare prediction models inspect forecasting metrics analyze demand trends