Characterizing On-Bus WiFi Passenger Behaviors by Approximate Search and Cluster Analysis

On-bus WiFi emerges as a promising market in recent years. It is an interesting problem to investigate the characteristics of bus passengers by the spatial-temporal data collected by smart WiFi devices. In this paper, we analyze passenger behavior logs, including WiFi connection, WiFi disconnection, web authorization, data traffic, etc. We aggregate these passengers' activities into online events by finding out the relations among these activities. Then we cluster Origin-Destination pairs of all the trips in a month of each passenger to find passenger's interested points and overall distribution of passengers' interested points. Our results show that there is only 8.33% of on-bus WiFi passengers to explore the web. The average time used on on-bus WiFi only last 6 minutes in one connection, but total time which a passenger spends on on-bus WiFi is about 25 minutes a day. On-bus WiFi passengers' average data traffic of the network is periodical. It is growing slowly until Sunday while dropping down on Monday, which indicates passengers prefer to use on-bus WiFi more frequently at weekends. 44.7% of the passengers are active in only one place and 39% of the passengers are active in two places.