Collaborative Filtering Recommendation Algorithms Performance on an Implicit Feedback Dataset
Name
Kristjan Lõhmus
Abstract
This thesis aims at investigating recommender systems based on a prior implicit dataset to enhance customer satisfaction on an online gaming platform operating in the United States (US). Such model-based algorithms as Alternating Least Squares (ALS) and Bayesian Personalized Ranking (BPR) were chosen in addition to baseline algorithms. A dataset including implicit preferences that was received from the platform was used to test and implement the models. The results showed an identical performance on evaluating the whole system, howev-er ALS performed better with players who were new to the system.
Graduation Thesis language
English
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Anna Aljanaki, Hakan Berber
Defence year
2021