Ensemble Convolutional Neural Networks for Mode Inference in Smartphone Travel Survey

We develop ensemble convolutional neural networks (CNNs) to classify the transportation mode of trip data collected as part of a large-scale smartphone travel survey in Montreal, Canada. Our proposed ensemble library is composed of a series of CNN models with different hyper-parameter values and CNN architectures. In our final model, we combine the output of CNN models using “average voting,” “majority voting,” and “optimal weights” methods. Furthermore, we exploit the ensemble library by deploying a random forest model as a meta-learner. The ensemble method with random forest as meta-learner shows an accuracy of 91.8% which surpasses the other three ensemble combination methods, and other comparable models reported in the literature. The “majority voting” and “optimal weights” combination methods result in prediction accuracy rates around 89%, while “average voting” is able to achieve an accuracy of only 85%.

I. Introduction

In the last two decades, advances in data collection techniques, particularly in the use of new technologies such as Global Positioning Systems (GPS) and smartphones have been changing the options available for collecting transportation demand data. Using the data collected from these emerging technologies typically requires methods of data processing and inference. Detecting transport mode has received the lion‘s share of attention in the literature as researchers have tried to infer mode with methods ranging from rule-based approaches to artificial intelligence algorithms.