Movie Recommender System – Tuning Asymmetric Singular Value Decomposition

I’m Winnie Nguyen, a senior majoring in Computer Science and Quantitative Economics at Earlham College.

About my CS Capstone Project:
As COVID-19 triggered stay-at-home orders and cancelled all social plans, people spend most of their time using online media streaming services. Among those streaming online providers, Netflix became the world’s leading Internet television network and the most-valued streaming service, which urges them to personalize users’ experience more correctly with the help of recommendation system. 

However, with the overload of vast amounts of customer data, recommender systems face challenges in processing data robustly and accurately. Moreover, missing values in matrix or sparsity is another challenge . In this paper, instead of using built-in function of Singular Value Decomposition (SVD)-based algorithms for recommender systems, I built Asymmetric SVD and tuned its hyper-parameters, expecting to make an improvement in prediction accuracy and running time.

Data Architecture Diagram

Senior Capstone Poster

Software Demonstration Video

Tuned Asymmetric SVD Paper

Tuned Asymmetric GitLab