Bus Arrival Prediction
Bus arrival prediction is a task of predicting the potential delay time of bus arrival. Accurate prediction provides users with an approach to more efficiently plan their day to day life with respect to the transportation and minimize their waiting time. In this report, we have applied 2 of the most popular machine learning techniques in this field, support vector regression (SVR) and artificial neural network (ANN), and compared their performance in bus arrival prediction based on features such as schedule time, whether it’s in rush hour or not, delay time of previous bus at current stop, delay time of target bus at previous stop, weather and weekday. We test those algorithms on 10 different datasets, one stop each, and use symmetric mean absolute percentage error (sMAPE) to evaluate and compare the performance of each model at each stop. Our experiment show that, with respect to our dataset, SVR models outperformed ANN models at most bus stops.