Design And Implement Of Context-Aware Recommender Devel- Opment Framework Based On Multidimensional Context

Context-aware recommender system utilizes context information to improve recom-mender accuracy. Traditional context-aware recommender algorithms do not focus on specific properties of context when integrating context into recommendation. Few of them can handle multidimensional context well. From the aspect of context-aware recommender development framework, some researchers just combine context-aware framework and recommender system together, and thus get heavyweight framework. Responding to these challenges, we propose a context pre-filtering recommender algo-rithm named ConRec based on multidimensional context. It takes full consideration of context properties that would influence the recommender process and recommender re-sults. It can also resolve the sparsity problem caused by multidimensional context well. Based on ConRec, we implement a lightweight development framework for context-aware recommender system. Lastly, we evaluate effectiveness and efficiency of Con-Rec through various experiments. The contributions can be summarized as follows:· By analyzing the problem that existing context-aware recommender systems did not focus on context information properties, we take the dynamic nature of con-text into full consideration from different aspects to get better recommendation result. We propose an approach to dynamically modeling context and personaliz-ing user’s preference on different kind of context. It can support context dynamic evolution as well.· We provide a context pre-filtering algorithm called ConRec based on multidi-mensional context. It handles the multidimensional context by decomposing data space and resolves the sparsity problem. ConRec also provides distributed algo-rithm for handling bigger data.· Based on ConRec, we implement a lightweight context-aware recommender sys-tem development framework, which can help define and configure context easily. It is built as a highly reusable and extensible software framework. Moreover it supports running on distributed environment.· We evaluate ConRec by taking experiments in single machine and distributed environment respectively. We compare it with other pre-filtering algorithms and conclude that ConRec gets more accurate recommendation.