Discriminative temporal interest-based preference learning for content-based recommendation
Content-based recommender systems (CBRSs) have shown to be highly effective for enhancing user experience in various real-world problems and application domains. This type of systems recommends items whose content is similar to the ones previously preferred by a user. Most existing CBRSs learn stati...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2021
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Content-based recommender systems (CBRSs) have shown to be highly effective for enhancing user experience in various real-world problems and application domains. This type of systems recommends items whose content is similar to the ones previously preferred by a user. Most existing CBRSs learn static user profiles assuming that user preferences remain stable over time. However, user preferences constantly change over time and, indeed, are attributed to a set of certain attributes or concepts at a certain period of time due to the existence of temporal dynamics in user preferences. Lacking the efficiency in handling these problems in CBRSs poses both semantical and temporal incoherence in user profiles due to the accumulated noise (i.e., incoherent attributes), which affects the accuracy of recommendation. This thesis presents Discriminate2Rec, a discriminative temporal interestbased content-based recommendation framework that employs a novel three-stage preference learning model that discriminates between items’ attributes based on their influence on user temporal preferences to improve both temporal and semantical attribute-level profile coherence for more accurate recommendation. We exploit different user-dependent and item/attribute-dependent temporal dynamics to infer positive and negative user-attribute temporal interest weights. Also, we introduce a negation modelling technique to model user-attribute negative interests, which allows us to learn attribute-level coherent user profiles. Evaluation is made on three real-world data sets. The results demonstrate the effectiveness of Discriminate2Rec in outperforming a variety of state-of-the-art methods in terms of recommendation accuracy. Furthermore, our three-stage preference learning model allows for significant improvements on the utilisation of signal and on the coherence of user profiles, leading to improve the recommendation accuracy consistently over time. |
|---|