Group-aware prediction with exponential smoothing for collaborative filtering

Abstract

In this paper, we describe our approach to the households recommendation track of Challenge on Context-Aware Movie Recommendation(CAMRa) 2011. The challenge of the track is to generate recommendations for households. In this paper, we introduce an approach that uses time information and exponential smoothing to predict user and item ratings. We provide traditional collaborative filtering algorithms as a baseline, and show that our approach yields better results.

A large-scale exploration of group viewing patterns

“…This dataset also provides household membership identifiers, but this "group" component is substantially smaller: it accompanies a user's rating for only 290 households. Many group recommendation approaches have been proposed and evaluated using this dataset [7,9,10]. Similarly, a large-scale dataset from the BARB organization is used in [19], which consists of about 15,000 users, 6,400 households, and 30 million TV program views.…”