This repository contains an academic project on bike rental demand prediction using a linear regression framework.
The project was conducted at Université Paris Dauphine – PSL as part of the Linear Model course, under the supervision of Katia Meziani.
The objective is to model and predict the number of bike rentals based on temporal, weather-related, and categorical variables.
A strong emphasis is placed on model interpretability, statistical rigor, and validation of linear regression assumptions.
The project follows a classical econometric workflow:
- Exploratory data analysis (EDA)
- Feature engineering and variable transformations
- Handling of categorical variables and interaction effects
- Linear regression modeling
- Hypothesis testing and ANOVA
- Model selection using AIC/BIC
- Diagnostic checks (normality, homoscedasticity, independence)
- Out-of-sample evaluation
The final model achieves strong predictive performance while remaining interpretable, making it suitable for operational decision-making in bike-sharing systems.
Detailed analysis, results, and discussions are provided in the accompanying report.
├── PROJET_MLG.pdf # Full project report (analysis & results)
├── R code for MLG project.Rmd # R Markdown source code