Collaborative Filtering Recommendation on Users’ Interest Sequences

Abstract

As an important factor for improving recommendations, time information has been introduced to model users’ dynamic preferences in many papers. However, the sequence of users’ behaviour is rarely studied in recommender systems. Due to the users’ unique behavior evolution patterns and personalized interest transitions among items, users’ similarity in sequential dimension should be introduced to further distinguish users’ preferences and interests. In this paper, we propose a new collaborative filtering recommendation method based on users’ interest sequences (IS) that rank users’ ratings or other online behaviors according to the timestamps when they occurred. This method extracts the semantics hidden in the interest sequences by the length of users’ longest common sub-IS (LCSIS) and the count of users’ total common sub-IS (ACSIS). Then, these semantics are utilized to obtain users’ IS-based similarities and, further, to refine the similarities acquired from traditional collaborative filtering approaches. With these updated similarities, transition characteristics and dynamic evolution patterns of users’ preferences are considered. Our new proposed method was compared with state-of-the-art time-aware collaborative filtering algorithms on datasets MovieLens, Flixster and Ciao. The experimental results validate that the proposed recommendation method is effective and outperforms several existing algorithms in the accuracy of rating prediction.

Content Curatorship and Collaborative Filtering: A Symbolic Interactionist Approach

“…The bulk of these studies appear in Information Systems journals and focus on the mechanics and possibilities of intelligent software (e.g. Barragáns-Martínez et al, 2010;Bobadilla, Ortega, & Hernando, 2012;Cheng, Yin, Dong, Dong, & Zhang, 2016;Liao & Lee, 2016;Nilashi, Jannach, Ibrahim, & Ithnin, 2015). Users are aggregated or clustered by their shared ratings of similar items, and various means of optimal aggregation are debated and tested (Bobadilla et al, 2012;Jeong, Lee & Cho, 2010;Lee, Lee, Lee, Hwang & Kim, 2016;Ortega et al, 2013Ortega et al, , 2016Xin, Ouuyang & Zhang, 2011).…”

“…Users are aggregated or clustered by their shared ratings of similar items, and various means of optimal aggregation are debated and tested (Bobadilla et al, 2012;Jeong, Lee & Cho, 2010;Lee, Lee, Lee, Hwang & Kim, 2016;Ortega et al, 2013Ortega et al, , 2016Xin, Ouuyang & Zhang, 2011). Tests have looked at the use of Pareto dominance (Ortega et al, 2013), clustering algorithms (Liao & Lee, 2016;Nilashi et al, 2015), semanticbased interest sequences (Cheng et al, 2016) and the recency and update of material (Jeong et al, 2010), to name but a few.…”

A User Interest Recommendation Based on Collaborative Filtering

“…Patra et al [10] propose a new similarity measure using Bhattacharyya coefficient for collaborative filtering in sparse data. Cheng et al [11] propose a new collaborative filtering recommendation method based on users' interest sequences (ISCF). These updated similarities, transition characteristics and dynamic evolution patterns of users' preferences are considered.…”

“…TJBCF is compared with CFBUI proposed in reference [9], ISCF proposed in reference [11] on Movielens datasets. It is depicted in Figure III that the MAE value of TJBCF is 3.4% lower than ISCF and 10.8% lower than CFBUI.…”

An improved memory-based collaborative filtering method based on the TOPSIS technique

“…In contrast, CF-based systems propose items based on an analysis of user feedback along with the preferences of similar users [3, 27–32]; this additional robustness makes CF the most widely used and successful RS method. CF approaches can be further classified into model- and memory-based techniques [1, 33–35]. Model-based approaches apply a pre-built model for predicting user preferences, whereas memory-based approaches (also known as neighbor-based models) access entire databases of user-provided ratings to find correlations between users/items.…”