detectBiklio: detecting bicycle usage with a smartphone app

The importance of cycling as an alternative mean of transportation and leisure is ever-increasing. In addition, the current ubiquity of mobile computing made it possible to be used to detect the transportation used. Thus, we present three possible solutions to detect cycling activity using a smartphone which are based on: Google Activity Recognition API, BLE sensors, and a Machine Learning classifier. An innovative and common aspect to all these solutions is that they do not require network connection to run, i.e., they execute solely on the smartphone. We present their architecture and evaluate them according to a set of requirements for the mobile app. Based on their evaluation, we concluded that the most interesting solution is based on machine learning given its balanced respect for the app requirements and possible evolution.