[4Rin1] Interactive 2

[4Rin1-88] Predicting spikes in cab ride time series using a deep neural network

〇Ayaka Yamamoto¹, Bojian Yang¹, Shin Ando¹ (1.Graduate School of Management, Tokyo University of Science)

Keywords:Deep Learning, Recurrent Neural Net, Time Series Prediction

This paper addresses the task of time series prediction on the number of cab rides embarking and disembarking within a designated area. The time series at hand contain spikes, the quantities of which are difficult to predict using regression models. We propose a hybrid deep neural network, comprising a convolutional subnet and LSTMs, for predicting the occurences or the absence of spikes during the commute hours. We conduct an empirical study using a real-world data and present a graphical application for guiding cab drivers based on the predicted numbers of rides.