Comparative Sequential Pattern Mining of Human Trajectory Data Collected from a Campus-wide BLE Beacon System

Abstract

Many social issues are expected to be addressed by collecting human trajectory data and analyzing them. As a demonstration study, we need a continuous and instant localization and trajectory collection system. We have developed a localization system using Bluetooth Low Energy (BLE) beacons and smartphones in our college campus. The system has been established to realize automated student roll call with 1, 600 BLE beacon emitters installed on our campus. We can estimate the location of a smartphone in our campus by analyzing received BLE beacons and their RSSIs (Received Signal Strength Indicators). In this paper, we demonstrate how we collect human trajectory data and how we can detect specific human behaviors from the collected data. We have obtained human trajectory data from 169 research participants comprised of 671 trips during the study held as a college festival event. Each research participant walked around with his/her smartphone. The smartphone continuously received BLE beacons during the event and periodically sent them to the server as a trajectory. We apply comparative sequential pattern mining to the obtained trajectory data and extract sequential patterns that are different between male trajectories and female trajectories. This study demonstrates the effectiveness of human trajectory data collection by a BLE beacon system and data analysis by comparative sequential pattern mining.