TRAVEL MODE PREDICTION

Machine Learning - 2025

This project explores how socioeconomic factors influence travel mode choice among Melbournians, using the VISTA 2023-2024 survey dataset. Working with a group, I helped build a data preprocessing pipeline that merged household, persons and travel datasets, handling missing values, outliers and feature engineering to create a unified dataset with over 6,000 households. We then trained k-nearest neighbours and a decision tree models to predict travel mode, evaluating performance through accuracy, F1 score and feature importance analysis. The decision tree offered clear interpretability, revealing the household location, income and age were strongest predictors of transport choice.