Online travel mode detection method using automated machine learning and feature engineering

Online travel mode detection provides context information useful for location-based services, in order to deliver a customized user experience. In the last years, many smartphone-based travel mode detection techniques have been proposed, but few explored the usage of dimensionality reduction in conjunction with hyperparameter optimization to improve accuracy with a reduced cost. In this paper, we propose a method to improve the accuracy and computational cost trade-off of travel mode detection, in which use state-of-the-art Feature Engineering and Automated Machine Learning techniques. In addition, we apply the proposed method in a real mobility dataset using different features and parameters. Our experiments showed that the combination of these techniques can greatly improve online detection performance. (C) 2019 Elsevier B.V. All rights reserved.