A Comparative Analysis of Multiple Algorithmic Techniques for the Development of an Intelligent System University Performance Evaluation based on CHED-DBM Ranking for selected SUCs

This study aims to compare the use of multiple algorithmic techniques for the development of an intelligent system University Performance Evaluation based on CHED-DBM Ranking for selected SUCs. The technique is based on the criteria of the CHED-DBM, which evaluate institutions' performance in Key Results Area (KRA). Evaluating the performance of SUCs is complex due to the multi-dimensional nature of their operations, such as instruction quality, research, extension services, equity, and governance. Various statistical and machine learning approaches are investigated to develop a solid evaluation model. It comprehensively compares multiple machine learning algorithms (Linear Regression, Decision Trees, Support Vector Machines (SVM), Random Forest and Feedforward Neural Network) in the development of an intelligent system and were evaluated using the essential performance metrics to determine the optimal method. The results demonstrate that Feedforward Neural Network have the best prediction accuracy (92%) and F1 score (89%), making them appropriate for complicated decision-making. However, their limited interpretability (1) and computational efficiency (1) could be improved. On the other hand, the Linear Regression and Decision Trees show high interpretability (5 and 4, respectively) and computational efficiency, although they have poorer predictive effectiveness. Random Forest and SVM achieve a balance of accuracy (90% and 88%, respectively) and efficiency while maintaining moderate interpretability. This study shows how the five (5) algorithms reveal the performance comparisons and the best methods for precise and scalable algorithm to be integrated in the development of the system.