Date of Award

2019

Publication Type

Master Thesis

Degree Name

M.Sc.

Department

Computer Science

First Advisor

Christie Ezeife

Keywords

Collaborative Filtering, Data Mining, E-commerce Recommendation, Historical Sequential Recommendation, Sequence Product Recommendation, Sequential Pattern Mining

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Abstract

In E-commerce recommendation system accuracy will be improved if more complex sequential patterns of user purchase behavior are learned and included in its user-item matrix input, to make it more informative before collaborative filtering. Existing recommendation systems that use mining techniques with some sequences are those referred to as LiuRec09, ChoiRec12, SuChenRec15, and HPCRec18. LiuRec09 system clusters users with similar clickstream sequence data, then uses association rule mining and segmentation based collaborative filtering to select Top-N neighbors from the cluster to which a target user belongs. ChoiRec12 derives a user’s rating for an item as the percentage of the user’s total number of purchases the user’s item purchase constitutes. SuChenRec15 system is based on clickstream sequence similarity using frequency of purchases of items, duration of time spent and clickstream path. HPCRec18 used historical item purchase frequency, consequential bond between clicks and purchases of items to enrich the user-item matrix qualitatively and quantitatively. None of these systems integrates sequential patterns of customer clicks or purchases to capture more complex sequential purchase behavior. This thesis proposes an algorithm called HSPRec (Historical Sequential Pattern Recommendation System), which first generates an E-Commerce sequential database from historical purchase data using another new algorithm SHOD (Sequential Historical Periodic Database Generation). Then, thesis mines frequent sequential purchase patterns before using these mined sequential patterns with consequential bonds between clicks and purchases to (i) improve the user-item matrix quantitatively, (ii) used historical purchase frequencies to further enrich ratings qualitatively. Thirdly, the improved matrix is used as input to collaborative filtering algorithm for better recommendations. Experimental results with mean absolute error, precision and recall show that the proposed sequential pattern mining-based recommendation system, HSPRec provides more accurate recommendations than the tested existing systems.

Share

COinS