Real Time Prediction of Taxi Services Using Deep Learning

Abstract

Modelling travel requirement is an essential part of transportation organisation. Searching passenger hotspots, balancing requirement and supply
problems, and re-allocating taxis to help drivers locate customers can all be aided by analysing demand. Predicting the number of taxi needed around the
city will help the taxi owners to coordinate the taxi fleet and reduce passenger and driver wait time. In this project, we propose a sequence learning model
that, based on recent requirement and other similar data, can predict the estimated taxi demand in each area of a region. Since future taxi requests are
associated with knowledge about past behaviour, recalling the data from the past is basic. By separating the city into more modest regions and foreseeing
the interest in each area, we evaluate our approach on a data set of taxi needed. We show that this method outperforms feed-forward neural networks and other prediction methods. We also display how other related data, such as time, temperature and drop-offs, influences the outcomes.