Towards Prediction of Bus Arrival Time using Multi-layer Perceptron (MLP) and MLP Regressor

Abstract

Intelligent transport systems have been in research and development in recent decades. However, not all countries can afford to deploy such systems for the public usage. Conventional public transport systems such as public buses are still the main mode of public transportation system in many developing countries. Due to the issue of public transportation's inaccurate bus arrival timing, the general public still prefers private transportation. The goal of this study is to investigate the use of machine learning to improve the prediction accuracy of bus arrival timing. Two machine learning models, a multi-layer perceptron (MLP) and a MLP regressor, were compared in terms of their performance on small datasets. The experiment data was collected from Kulai-Johor Bahru Sentral bus route in Malaysia and cleaned to negate errors that influenced the accuracy of the models. The performances of the models were analysed and discussed and we observed that the MLP outperforms the MLP regressor. A limitation of this study is the small dataset that only comprises bus location data collected on a single bus route.