Implementasi Metode Content-Based Filtering untuk Rekomendasi Soal Berbasis Performa User dalam Mendukung Personalized Learning pada Website LoLosASN

Abstract

A recommendation system is an effective approach to supporting more adaptive learning processes, particularly in helping users prepare for the CPNS (Indonesian civil servant) exam. This study aims to develop a question recommendation system based on user performance using the Content-Based Filtering (CBF) approach on the LolosASN website. The system is designed to analyze users’ weaknesses based on their question-solving history and recommend practice questions that are relevant to those weak areas. Question representation is carried out using the Term Frequency-Inverse Document Frequency (TF-IDF) method, while the matching between the user profile and item profile is calculated using cosine similarity. The user profile is constructed from subtopics with an error rate of ≥ 70% and is prioritized in the recommendation process. The system was tested under three parameter combination scenarios, with the best results obtained using the configuration of n-gram (1,2) and max_df of 0.2, achieving a precision of 0.9656 and an nDCG score of 0.9889. Additionally, a user evaluation conducted through the ResQue questionnaire showed high user satisfaction, with an average score above 4.5 out of 5 across all dimensions. These results indicate that the developed recommendation system not only provides questions that are relevant to users’ weaknesses but also delivers a high-quality interaction and user experience. Thus, this system makes a significant contribution to supporting the implementation of more effective personalized learning, especially in the context of online learning for competitive exam preparation such as the CPNS.