Enhancing Programming Competition Performance: A Data-Driven Approach to Personalized Training

In programming competition training, the key to boosting students' competitiveness and award chances is to provide personalized problem recommendations based on their coding skills. Traditional uniform training problems fail to account for the individual differences among students, and cannot guarantee a perfect alignment between problems and students' abilities. Each student possesses a unique learning pace, comprehension level, and mastery degree, making a standardized method likely to overlook or excessively chal-lenge certain students. This impacts their learning effective-ness and enthusiasm. The personalized and flexible problems recommendation algorithm is urgently needed. We propose four novel problem recommendation algorithms, which aim to facilitate the personalized generation of problem sets tailored to students' skill levels. Specifically, we use programming history records to evaluate students' strengths and weaknesses, to recommend suitable exercises and adjust recommendation algorithms based on their performance in these programming problems. In a case study at Beijing University of Chemical Technology (BUCT), the implementation of these personalized training algorithms leads to a 33% improvement in final performance compared to traditional approaches. This suggests a promising new direction for programming education.