MSc Thesis Defense Announcement of Md Burhan Uddin:"Multi-Data Source Recommendations with Derived Sequential Pattern Mining"

Friday, September 9, 2022 - 10:00 to 12:00

The School of Computer Science is pleased to present… 
MSc Thesis Defense by: Md Burhan Uddin 
Date: Friday September 9, 2022 
Time:  10:00am - 12:00pm  
Location: Lambton Tower, Room 3105 
Reminder: All attendees must record the two-part attendance; QR Code online attendance sheet AND the attendance sign-in sheet.


     User-based Collaborative Filtering (CF) methods fail to derive preference values for new items with no ratings or purchase information. Also, a CF system's performance depends on the user-item matrix's sparsity. Using multi-storage solutions (e.g., multiple related tables in a relational database, distributed and multi-platform databases) in E-commerce and attribute information of items provides the opportunity to improve the recommendation process and accuracy through multi-source data integration. For example, with a traditional Recommendation System (RS), integrating purchase sequence similarity helps find users with similar interests in resolving sparsity issues. Using the item attributes information (e.g., category) to derive user preferences on an item helps generate a recommendation for similar items with no rating or purchase history, alleviating the cold start problem.
     Existing multi-source RS algorithms, such as Extended Category-based Collaborative Filtering (ECCF18), address the cold start problem using item categories to extend the CF model and enhance the preference matrix to derive recommendations. ECCF18 does not consider user purchase pattern history and fails to generate personalized recombination for users with few to no ratings. The Cross Domain Sequential Pattern Mining (CD-SPM19) model uses item ontology to derive semantic similarity information to enrich the collaborative filtering (CF) step. The main limitation of the CD-SPM19 model is that it fails to capture changes in user preference over time. Both algorithms fail to learn complex sequential relations in user purchase behavior without utilizing sequential purchase information. The Historical Purchase and Clickstream Recommendation (HPCRec18) model derives a consequential bond between click to purchase to predict preferences for users with no purchase history but fails to capture the sequential relations. Finally, the Historical Sequential Pattern Recommendation (HSPRec19) model using historical sequential purchase database (SHOD) algorithm derives sequential purchase patterns from historical purchase data. It then uses the consequential bond in mined patterns to improve the user-item matrix for CF. Both HPCRec18 and HSPRec19 cannot derive recommendations for new item if the item has no click or purchase data in the historical transaction database. 
     This thesis proposes the Multi-source Category Extended Historical Sequential Pattern Recommendation System (MCE-HSPRec), an extension of the HSPRec19, which uses item categories to enrich user profiles and address the new item problem. The proposed algorithm derives enriched category-based user profiles by analyzing purchase behavior and item categories that are frequently purchased together. MCE-HSPRec first generates a rich user-item matrix (UI) using sequential relation derived from sequential pattern mining in the HSPRec module. Then it derives a Category Co-occurrence Graph (CCG) and a user category preference matrix (UC) from the combination of historical e-commerce and item category information. The thesis computes a category-based user profile using the UI, CCG, and UC before applying CF. The final user-category preference matrix is much denser, thus improving recommendation performance. Experimental results show that the proposed MCE-HSPRec outperforms existing systems providing more accurate prediction and increasing item prediction coverage in the domain. 
Keywords: Multi-Source, Sequential Pattern Mining, Category Co-occurrence, Recommendation System 

MSc Thesis Committee:  

Internal Reader:                Dr. Yung H. Tsin 
External Reader:               Dr. Scott Mundle 
Advisor:                             Dr. Christie Ezeife 
Chair:                                 Dr. Boubakeur Boufama

MSc Thesis Defense Announcement

Vector Institute in artificial intelligence, artificial intelligence approved topic logo

5113 Lambton Tower 401 Sunset Ave. Windsor ON, N9B 3P4 (519) 253-3000 Ext. 3716