The University of Windsor is preparing for a safe return to campus. Learn More.

MSc Thesis Defense Announcement of Hemni Sri Rajeswari Karlapalepu:"A Taxonomy of Sequential Patterns Based Recommendation Systems"

Friday, July 31, 2020 - 10:00 to 12:00


The School of Computer Science is pleased to present…

MSc Thesis Defense by:  Hemni Sri Rajeswari Karlapalepu

Date:  Friday, July 31st, 2020
Time:  10:00 am – 12:00 pm


With remarkable expansion of information through the internet, users prefer to receive the exact information they need through some suggestions to save their time and money. Thus, recommendation systems have become the heart of business strategies of E-commerce as they can increase sales and revenue as well as customer loyalty. Recommendation systems techniques provide suggestions for items/products to be purchased, rented or used by a user. The most common type of recommendation system technique is Collaborative Filtering (CF), which takes user’s interest in an item (explicit rating) as input in a matrix known as the user-item rating matrix, and produces an output for unknown ratings of users for items from which top N recommended items for target users are defined. E-commerce recommendation systems usually deal with massive customer sequential databases such as historical purchase or click sequences. The time stamp of a click or purchase event is an important attribute of each dataset as the time interval between item purchases may be useful to learn the next items for purchase by users. Sequential Pattern Mining mines frequent or high utility sequential patterns from a sequential database. Recommendation systems accuracy will be improved if complex sequential patterns of user purchase behavior are learned by integrating sequential patterns of customer clicks and/or purchases into the user-item rating matrix input. Thus, integrating collaborative filtering (CF) and sequential pattern mining (SPM) of historical clicks and purchase data can improve recommendation accuracy, diversity and quality and this survey focuses on review of existing recommendation systems that are sequential pattern based exposing their methodologies, achievements, limitations, and potentials for solving more problems in this domain.
This thesis provides a comprehensive and comparative study of the existing Sequential Pattern-based E-commerce recommendation systems (SP-based E-commerce RS) such as ChoRec05, HuangRec09, LiuRec09, HOPE RecSys12, Hybrid Model RecSys16, Product RecSys16, HPCRec18 and HSPCRec19. Thesis shows that integrating sequential patterns mining (SPM) of historical purchase and/or click sequences into user-item matrix for collaborative filtering (CF) (i) Improved recommendation accuracy (ii) Reduced limiting user-item rating data Sparsity (iii) Increased Novelty Rate of the recommendations and (iv) Improved Scalability of the recommendation system. Thus, the importance of sequential patterns of customer behavior in improving the quality of recommendation systems for the application domain of E-commerce is accentuated through this survey by having a comparative performance analysis of the surveyed systems. 
Keywords: sequential patterns, frequent patterns, sequential pattern mining, e-commerce, recommendations, recommender systems, collaborative filtering, clickstream history

Thesis Committee:

Internal Reader: Dr. Dima Alhadidi
External Reader: Dr. Dilian Yang
Advisor: Dr. Christie Ezeife
Chair: Dr. Saeed Samet

MSc Thesis Defense Announcement


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