Machine learning for new friends recommendation in NTU

Abstract

NiceToMeetU (NTmU) was developed as a friend recommendation system to help undergraduate students in Nanyang Technological University (NTU) to make new friends on campus. While the project is a continuous effort, my work involves mainly creating a software application for users. Previous work aimed to improve the matching algorithm using machine learning, which included: - Construction of user-profiles through questionnaires, which helped to collect textual data from users. - Deep learning approaches such as the Bidirectional Encoder Representations from Transformers (BERT). - Web application designed to collect more featuring data from our users including movies and music. My FYP focuses on delivering a user-friendly mobile application that encourages user usage of the NTmU platform. While there was initially a previous mobile application built, it still lacked certain functionality like the forum and chat. As such, the opportunity to leverage these functionalities as tools to collect textual data, as well as a more compelling reason to increase their screentime on NTmU, was missed out on.