Research Of Collaborative Filtering Based Context Prediction

Backup by the advancing of mobile computing technology, pervasive computing has become a research hotspot. The key tasks of a pervasive computing are gathering, processing and feedback of information. Context-aware computing provides the core functionalities to solve fulfill this set of requirements. Thus, it becomes central important for us to look closely at context-aware computing.This paper focuses on the research of context prediction in a context-aware computing environment. We aim to provide an ontology-based context model with a hierarchal structure as the foundation of context prediction. Our context prediction method leverages the advantages of our context model. We integrate collaborative filtering, which is widely used in the e-commerce literature, into the context prediction process. We have adopted methods that integrate the similarity of context into the prediction process. And besides, to overcome the slow start problem of the context-aware collaborative filtering process, we proposed a prediction method based on the sub-class of the user base according to user's interest.The major contributions of the paper are:1. Integrating collaborative filtering into context prediction process, overcoming the slow-start problem by combining user sub-classing and content-based prediction.2. Designed and implemented a prototype middleware system based on our proposed prediction method. An practical application demonstration is also provided.