Mining association rules with multiple minimum supports
KDD '99 Proceedings of the fifth ACM SIGKDD international conference on Knowledge discovery and data mining
Effective personalization based on association rule discovery from web usage data
Proceedings of the 3rd international workshop on Web information and data management
Efficient Adaptive-Support Association Rule Mining for Recommender Systems
Data Mining and Knowledge Discovery
Fast Algorithms for Mining Association Rules in Large Databases
VLDB '94 Proceedings of the 20th International Conference on Very Large Data Bases
Robustness of collaborative recommendation based on association rule mining
Proceedings of the 2007 ACM conference on Recommender systems
Accuracy improvements for multi-criteria recommender systems
Proceedings of the 13th ACM Conference on Electronic Commerce
ISMIS'12 Proceedings of the 20th international conference on Foundations of Intelligent Systems
Fuzzy association rule mining approaches for enhancing prediction performance
Expert Systems with Applications: An International Journal
An improved neighborhood-restricted association rule-based recommender system
ADC '13 Proceedings of the Twenty-Fourth Australasian Database Conference - Volume 137
Hi-index | 0.00 |
Association rule mining algorithms such as Apriori were originally developed to automatically detect patterns in sales transactions and were later on also successfully applied to build collaborative filtering recommender systems (RS). Such rule mining-based RS not only share the advantages of other model-based systems such as scalability or robustness against different attack models, but also have the advantages that their recommendations are based on a set of comprehensible rules. In recent years, several improvements to the original Apriori rule mining scheme have been proposed that, for example, address the problem of finding rules for rare items. In this paper, we first evaluate the accuracy of predictions when using the recent IMSApriori algorithm that relies on multiple minimum-support values instead of one global threshold. In addition, we propose a new recommendation method that determines personalized rule sets for each user based on his neighborhood using IM-SApriori and at recommendation time combines these personalized rule sets with the neighbors' rule sets to generate item proposals. The evaluation of the new method on common collaborative filtering data sets shows that our method outperforms both the IMSApriori recommender as well as a nearest-neighbor baseline method. The observed improvements in predictive accuracy are particularly strong for sparse data sets.