Investigating Feature Selection for Predicting the Number of Bus Passengers by Monitoring Wi-Fi Signal Activity from Mobile Devices

Abstract

The quality of service on public transportation system is motivated by passengers. It can be improved in several ways, such as bus arrival on time and seats available on the bus. Our research goal tries to provide congestion information to passengers in a bus transportation system by predicting the number of passengers. This can be done using several techniques. However, the most important factor is the prediction model, which consists of features and an algorithm. In this paper, we focus on feature selection using the Recursive Feature Elimination (RFE) method which can improve the prediction accuracy. We also show our experimental results by comparing the actual observed number of passengers at the bus stop to the predicted number of passengers.