iHSPRec: Image Enhanced Historical Sequential Pattern Recommendation - Abdulrauf Gidado

Abdulrauf Gidado presents "iHSPRec: Image Enhanced Historical Sequential Pattern Recommendation" at the School of Computer Science Colloquium Series at the University of Windsor on February 2, 2024.

Abstract:
Collaborative filtering (CF) faces issues such as sparsity and cold start problems, while content-based approaches suffer from overspecialization and lack of personalization. Systems such as HSPRec18, HPCRec19, and SemRec have not explored using item images to enrich recommendations. However, cross-regional e-commerce systems like Amazon and eBay have retailers from disparate backgrounds. For instance, identical items are likely to be named or described differently by different retailers on these systems, such as “Slippers”, “Flipflop” and “Sandals”, or “Pad” as a female hygiene product or “Pad” as a computer mouse pad. The high image retrieval time of images stored on a relational database or as a file makes it difficult to incorporate images with item names and descriptions into recommendation input. To improve recommendation accuracy, this paper proposes iHSPRec (Image Enhanced Historical Sequential Pattern Recommendation), which stores item images in textual formats of Base64 or BLOB on a document-oriented NoSQL database. It incorporates items' image similarity scores to build a vectorized item-item similarity matrix and sequential pattern of users' purchases or clicks as input data for the recommendation process. This is achieved by learning the similarity between item images using the Structural Similarity Index (SSI) score, mining pictorially similar sequential purchase patterns, and enriching the item-item matrix with the products' structural similarity index score and sequential product purchase patterns. iHSPRec provides Top-K personalized recommendations based on image similarities between items without needing a user's rating. Experimental results and comparison with existing systems show that iHSPRec has higher recommendation accuracy than existing benchmark methods.

Biography:
Abdulrauf Gidado is a PhD Candidate at the School of Computer Science, University of Windsor in the domain of Bigdata and data mining with a research focus on Mining for Product Recommendations on Document-Based NoSQL Bigdata. He has been a recipient of various scholarships including three terms Mitacs Award, and the University of Windsor scholarships. Abdulrauf has also published several papers in ACM and IEEE conferences and journals. Abdulrauf is also performing responsibilities as a Sessional Instructor at the University of Windsor, teaching undergraduate students in Computer Science. Abdulrauf is passionate about teaching and software development and his teaching and industry career spans over a decade including in Africa, Europe and North America. During his career, he has had the opportunity to research and develop more than thirty enterprise solutions. This is in addition to organizing and attending various conferences and exhibitions for students and had led many training sessions, workshops, and competitions. Besides research, teaching and software development, Abdulrauf is actively involved in various mentoring activities. In recent years, he was part of various initiatives such as summer coding sessions for black youths in Windsor and Academic Excellence Initiatives (AEI). In his free time, Abdulrauf enjoys reading books and playing computer games.