Valid context detection based on context filter in context-aware recommendation system
Shulin Cheng, Bofeng Zhang, Guobing Zou, Mingqing Huang, Ying Da Lv · 2018
Context as a kind of quite important information plays a significant role in context-aware recommendation system (CARS). Many studies have been proved that context help promote to improve the effectiveness of recommendations. But a serious challenge has yet not been solved well, which is how to detect valid contexts for users in CARS, since different users have different sensitivity to contexts. Motivated by the observations, we proposed a method of valid context detection based on context filter. Context filter comprises two selection phases. In the first phase, context selection depends on the expert experiences, which is also called primary selection. We focus on the second selection phase named refinement selection based on one-way analysis of variance (ANOVA). By one-way ANOVA, a utility function is put forward to measure user's context sensitivity to detect valid contexts. We verified the effectiveness of detection method by the experiments on a small real film dataset.