Transportation Mode Recognition based on Cellular Network Data

A variety of modern technologies leveraging ubiquitous mobile phones have addressed the transportation mode recognition problem, which is the identification of how users move about (walking, cycling, driving a car, taking a bus, etc). It has found applications in areas such as smart city transportation, greenhouse emission calculation, and context-aware mobile assistants. To date, significant work has been devoted to the recognition of mobility modes from the GPS and motion sensor data available on smartphones. However, these approaches often require users to install a special mobile application on their smartphone to collect the sensor data, they are power inefficient and privacy intrusive. Also, bus and car modes are similar in terms of motion due to the same roads, traffic regulations and speed, which makes it challenging to accurately distinguish between the two modes. In this research, we handle these issues by offering a user-independent system that distinguishes three forms of locomotion-walk, bus, and car-solely based on mobile data (3G and 4G) of a smartphone. The system was developed using data collected in Makkah city, Kingdom of Saudi Arabia. Using statistical classification and boosting techniques, we successfully achieved an accuracy of 82%, and with post-processing step we achieved an accuracy of 99%.