| Summary: | 碩士 === 國立中興大學 === 資訊科學與工程學系 === 103 === In recent years, Location-based Reminder (LBR) app attracts significant attention due to its flexibility on reminding to-do things based on user’s location rather than fixed rigid time point. While the LBR apps become popular and common, we argue that the existing LBR apps suffer from the following two limitations. First, existing LBR apps mainly rely on Global Positioning System (GPS) to acquire user’s positions. However, using GPS sensors on mobile devices consumes lots of power. The high power consuming problem raise the availability concerns for LBR apps. Second, using GPS sensors will need the help of external service to convert GPS longitude coordinates and latitude coordinates into logical locations such that people can understand it.
Due to the disadvantage of GPS, in this study, we propose to build LBR apps by employing Wi-Fi signals. Our basic idea is based on the following observation: A Wi-Fi hotspot’s network name is often with location semantics and sometime highly related to the location where the Wi-Fi hotspot installed. For example, 7-11 convenience stores provided Wi-Fi service with a network name: 7-11Wi-Fi. By employing such observations, we can build Wi-Fi based LBR without the limitations of using GPS sensors. While such observation looks neat, we find that not all Wi-Fi network names with location semantics. Actually, by preliminary data collection and examination, we find that the portion of Wi-Fi network names having location semantics is few. How to select Wi-Fi network names having location semantics from thousands of Wi-Fi logs becomes a research challenge. In this paper, we start from the angle of machine learning to address the problem of selecting Wi-Fi network names having location semantics. Experiments using real data are conducted, and the results demonstrate the effectiveness of the proposed techniques.
|
|---|
