In educational programming courses, students are tasked with solving large-scale programming projects to apply and reinforce their understanding of key concepts and skills. Testcase Recommendation Systems (TRS) offer personalized support by suggesting curated testcases tailored to the skill levels of individual learners. However, existing approaches, such as those based on Regularized Singular Value Decomposition (RSVD), struggle to effectively incorporate students’ submission histories, limiting the personalization and diversity of recommendations. To address these challenges, this paper presents a novel Multi-model Testcase Recommendation System (MTRS) that integrates multiple recommendation models, including the enhanced TimeSVD model, which leverages temporal data to improve accuracy and adaptability. The proposed MTRS evaluates user satisfaction through two metrics: Rating and Usage, while assessing recommendation efficacy with Coverage and Novelty. It has also been successfully implemented in university-level programming courses, with feedback collected through detailed surveys. The results show that TimeSVD excels in generating testcases with high Novelty, whereas RSVD performs better in other metrics, demonstrating the complementary strengths of these models.