iDiary : compression, analysis, and visualization of GPS data to predict user activities

Abstract

"What did you do today?" When we hear this question, we try to think back to our day's activities and locations. When we end up drawing a blank on the details of our day, we reply with a simple, "not much." Remembering our daily activities is a difficult task. For some, a manual diary works. For the rest of us, however, we don't have the time to (or simply don't want to) manually enter diary entries. The goal of this thesis is to create a system that automatically generates answers to questions about a user's history of activities and locations. This system uses a user's GPS data to identify locations that have been visited. Activities and terms associated with these locations are found using latent semantic analysis and then presented as a searchable diary. One of the big challenges of working with GPS data is the large amount of data that comes with it, which becomes difficult to store and analyze. This thesis solves this challenge by using compression algorithms to first reduce the amount of data. It is important that this compression does not reduce the fidelity of the information in the data or significantly alter the results of any analyses that may be performed on this data. After this compression, the system analyzes the reduced dataset to answer queries about the user's history. This thesis describes in detail the different components that come together to form this system. These components include the server architecture, the algorithms, the phone application for tracking GPS locations, the flow of data in the system, and the user interfaces for visualizing the results of the system. This thesis also implements this system and performs several experiments. The results show that it is possible to develop a system that automatically generates answers to queries about a user's history.