An MCTS-Based Recommender System for Education Complex
Debin Zhao, Zhengyuan Hu, Yinjian Yang · 2022
Although recommender systems in the education industry filter overload information and bring convenience to both learners and education facilities, challenges such as the coldstart problem and issue of timeliness, however, are more significant than recommending traditional goods while providing educational advice for students. To address this issue, this paper proposes an innovative recommender system based on Monte Carlo tree search (MCTS), which performs root cause analysis of the increased number of students during a certain period in each education facility, followed by computation of similarities between root cause groups and the rest of the students using Mahalanobis Distance. An experiment with a six-month dataset obtained from an education complex named Luoyang City Hall was conducted, and the results of A/B testing indicate that the treatment group experiences a higher student attendance rate than samples using traditional recommendation methods.