Personalized information recommendation model based on context contribution and item correlation
•We propose an improved Association Rule method for mining user’s behavior pattern.•Measurement of user interest model combining with contextual preference.•We assess the applicability of Association Rule to item-based Collaborative Filtering.•Several set of experiments show the efficiency of these...
Gespeichert in:
Veröffentlicht in: | Measurement : journal of the International Measurement Confederation 2019-08, Vol.142, p.30-39 |
---|---|
Hauptverfasser: | , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | •We propose an improved Association Rule method for mining user’s behavior pattern.•Measurement of user interest model combining with contextual preference.•We assess the applicability of Association Rule to item-based Collaborative Filtering.•Several set of experiments show the efficiency of these three methods.•Fusion measurement of context and item information achieve better results.
The traditional information recommender system gives little consideration to the influence of contexts on users and correlations between items, thus affects the quality of personalized information recommendation service. In order to address the issue, the main contribution of this paper is to propose the measurement for context contribution and item correlation, and designs a novel personalized information recommendation model. First, the paper proposes an improved association rule algorithm based on FP-Tree to improve efficiency of user’s behavior pattern mining in big-data environment. Second, the paper puts forward a user interest extraction algorithm based on improved FP-Tree and context contribution to measure and model user preferences in personalized information recommendation service. Third, the paper presents an improved collaborative filtering algorithm based on item correlation. In order to deal with data sparseness, an “item-item” matrix is constructed by using frequent itemsets in association rules. Then, the paper uses context contribution to replace item score, which can improve the accuracy of the similarity measurement between items. Experiments show that the model is more effective and accurate than other existing methods. |
---|---|
ISSN: | 0263-2241 1873-412X |
DOI: | 10.1016/j.measurement.2018.12.004 |