Active Caching for Recommender Systems

Web users are often overwhelmed by the amount of information available while browsing and searching information. Recommender systems substantially reduce the information overload by suggesting a list of similar documents that users might find interesting. However, generating these ranked lists require enormous amount of resources and often results in access latency. Caching of frequently accessed data has been shown to be a useful technique in related areas for reducing stress on limited resources and improving response time. However, the traditional passive Caching techniques can only achieve performance gain up to a certain level and there is room for further improvement. In this research, we propose “Active Caching” techniques for content based recommender systems. We have proposed solutions for different conditions. These conditions vary by the type of distance measure used and state of access to information source. Our proposed methods not only answer queries that already exist in the cache, but also actively process queries that require aggregation or other computation on the data stored in the cache. These methods effectively provide solution for situations, where access to the data source is available or not. We also investigated the use of metric as well as non-metric distance functions. Through a simulation environment, using real dataset from an experimental site, we achieved substantial cache hit ratio when compared when passive caching approach.