Web-based matching system for friend recommendation

• Main Author: Ang, Shu Hui
• Other Authors: Hui Siu Cheung
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/166689
• Institution: Nanyang Technological University

NiceToMeetU (NTmU) is developed as a friend recommendation platform using Django Framework for Nanyang Technological University (NTU) students to make friends in school. The platform makes use of the content-based filtering recommendation system to match users based on their profiles, which are constructed from 4 aspects - personality traits, movies, hobbies, and music in the ratio of 4:3:2:1. The previous version classified users into 16 Myers Briggs Type Indicator (MBTI) personality type across 4 axes using BERT model. This project focused on improving the accuracy of personality classification and enhancing the user interface. Various word embedding models, such as BERT, RoBERTa, DistilBERT and ALBERT, were experimented with a new, larger dataset for personality classification. The best classification accuracy was obtained by the BERT model with a ROC AUC score of 91% and f1 score of 89%.