Artificial Intelligence driven optimal route planning for urban transit

This thesis proposes an ensemble approach by combining machine learning and deep learning techniques for predicting taxi trips arrival times and predicting shared rides. This research work considers the capabilities of machine learning and deep learning models to predict arrival times and shared rides as an essential aspect of developing an intelligent transportation system offering optimized and efficient ride-sharing schemes. The dataset was made publicly available by the New York City transport, and limousine company (TLC) consisting of roughly 1 million records of New York City green taxi trip data for January, February 2017 and January to June 2019. We compare the results of our ensemble approach and observe that the Linear regression model and Convolutional Neural Network (CNN) perform admirably for arrival time prediction while the Capsule Network (CapsNet) provides the best results for ride-sharing prediction. Keywords—Arrival time prediction; Ride-sharing prediction, Machine learning; Deep learning, Ensemble techniques.