The vast collection of movement data through various mobile devices generates a significant and precious amount of information that can provide valuable insight into movement behaviors in various applications. Researchers from different communities have developed models and techniques for mobility analysis, but they mainly are focused on the geometric properties of trajectories. As such, the techniques are good at discovering patterns, but the patterns are difficult to interpret in a particular application domain.
This thesis proposes an ontology based semantic knowledge discovery framework to understand mobility data and semantically interpret trajectory patterns. It consists of three main parts, namely: semantic trajectory ontology modelling, activity recognition, and semantic behavior modelling. First, a semantic conceptual data model is defined, which helps in developing an ontology model. The activity recognition part consists of several steps, namely: data preparation, semantic enrichment process, and semantic features extraction. Activity types were defined as axioms based on the semantic features. The retrieved information from the previous steps are used to populate the ontology model for classifying different activity types by reasoning. Next, the association rule mining algorithm, apriori, is applied to extract different behavior patterns. The process considers four different behavior types, namely: semantic, semantic and space, semantic and time, and semantic and space-time.
A system prototype was developed to evaluate the performance of the framework using a simulated dataset and two different real datasets. For evaluating the activity recognition model, the inferred activity types were compared to the activity types declared by users in the feedback. For evaluating the ontology based behavior model, one of the location based services named location based advertisement was considered to test different aspects of the extracted behavior models. In the activity recognition model, it was observed that the accuracy of the results was related to the availability of the points of interest around the places that users had stopped at some parts. In the semantic behavior modelling, the results showed that applying the extracted behavior rulesets could filter relevant services for the users from the number of available services and customize the services based on the rulesets.