AI Driven for Incoming Students Field Recommendation in Rwanda Polytechnic

This paper presents a machine learning-based recommendation system to optimize field placement for first-year learners entering Rwanda Polytechnic (RP) from both general education (REB) and Technical Secondary School (TSS) backgrounds. The study addresses the challenge of admitting students to technical programs that align with their academic performance to minimize dropout rates and module retake rates. Using a dataset of 399 admission records from two of eight RP colleges for the academic years 2024-2025 and 2025-2026, we examined 128 independent features (Board features: 2, Combination features: 22, Subject features: 104) derived from national examinations to predict the optimal field placement across 13 technical fields, including Information Technology, Civil Engineering, Quantity Surveying, and Irrigation and Drainage Technology. Three machine learning models were trained and tested, where an Artificial Neural Network (ANN) with 27,758 trainable parameters achieved 94.36% training accuracy and 94% testing accuracy, demonstrating a balanced fit and strong generalization performance. Entrepreneurship, applied technical courses, and language and communication subjects were identified as the most predictive features of student achievement across technical fields. The model identifies the most suitable fields for students and presents those who are likely to succeed in their admitted field. This research introduces a data-oriented methodology to enhance the admissions process, supporting students and academic advisors in making informed decisions. By promoting high-quality, skill-oriented education, it contributes to RP's mission of strengthening Technical and Vocational Education and Training (TVET) in Rwanda.