Predicting Traffic Volume in New Zealand with Machine Learning

As part of my 2 week capstone project, I worked on time series forecasting for predicitng the traffic volume for a given point in New Zealand using real world data obtained from the NZTA. The application of predicting traffic count for a given place in NZ, allows various stakeholders such as Transporation agencies to plan road maintenence, traffic signal timing, and new infrastructure based on expected traffic demand.

Github

Project Overview

This project aims to develop a predictive model for accurately forecasting traffic volume at a specific location and time. Historical traffic data and relevant features like date, time, day of the week, month, and weather conditions are utilized. Machine learning techniques are employed to capture patterns and trends in the data, enabling accurate predictions of future traffic volumes. The best-performing model is the random forest regression, which outperformed other models in all evaluation metrics. However, predicting traffic volume in 2018 proved challenging due to irregular data, highlighting the need for further improvements in data consistency and comparability.