Recommender System of Final Project Topic Using Rule-based and Machine Learning Techniques

The final project is a mandatory graduation requirement for bachelor's degree students. However, students often having problems in determining topic that is suitable with their interests and competencies. As a result, some students might have to change their topics halfway, which can affect their study period. Ironically, the abundant volume of previous final project documents available in the university library only add more confusion and difficulty for the students in finding relevant references for their research topic. Therefore, the focus of this research is to implement a machine learning approach to analyze and model an algorithm to recommend final project topics, based on student's interest, competencies, and their respective supervisor. This research also aims to establish a framework to map academic attributes, as part of feature selection. As the result, we develop a recommender system based on cosine similarity algorithm to recommend topics based on similarity between student's profile and topics represented by lists of keywords. Performance is measured by comparing the recommendations given by the proposed system against the actual topic chosen by students, with a very satisfying result of 71.43% precision.