A Comparison among various Classification Algorithms for Travel Mode Detection using Sensors’ data collected by Smartphones

Nowadays, machine learning is used widely for the purpose of detecting the mode of transportation from data collected by sensors embedded in smartphones like GPS, accelerometer and gyroscope. A lot of different classification algorithms are applied for this purpose. This study provides a comprehensive comparison among various classification algorithms on the basis of accuracy of results and computational time. The data used was collected in Kobe city, Japan using smartphones and covers seven transport modes. After feature extraction, the data was applied to algorithms namely Support Vector Ma-chine, Neural Network, Decision Tree, Boosted Decision Tree, Random Forest and Naïve Bayes. Results indicated that boosted decision tree gives highest accuracy but random forest is much quicker with accuracy slightly lower than that of boosted decision tree. Therefore, for the purpose of travel mode detection, random forest is most suitable. _______________________________________________________ M. A. Shafique (Corresponding author) • E. Hato Department of Civil Engineering, The University of Tokyo, 3-1-7, Hongo, Bunkyo-ku, Tokyo 113-8656, Japan Email: [email protected] Email: [email protected] CUPUM 2015 175-Paper