A Machine Learning Approach to Estimate Public Transport Ridership Using Wi-Fi Data

Obtaining accurate data on bus ridership is a challenge for public transport operators. Wi-Fi data, collected from sensors placed on buses, seem promising for generating O-D matrices over a network. However, the large amount of passive data obtained does not necessarily lead to a more accurate understanding of mobility patterns. Problems of completeness remain, as Wi-Fi sensors do not detect all signals emitted by connected devices in their vicinity, and some people do not own these devices. Data correction is therefore a crucial step in the process of building O-D matrices from Wi-Fi data. In recent years, the democratization of machine learning and artificial intelligence has provided interesting methodological avenues. In this work, machine learning algorithms (Random Forest, Light Gradient Boosting Machine and Multilayer Perceptron) are used to estimate the absolute values of passengers boarding and alighting at a bus stop based on Wi-Fi data. The main advantage of these models is their ability to use spatial and temporal features, which are necessary to overcome the problem of data completeness. The results show that LGBM is the most relevant algorithm for generating accurate data on the number of passengers boarding and alighting at each bus stop, taking into account their spatio-temporal variability.