Prediction Analysis of Students’ Performance on the Hybrid EDA-SVR Model

Abstract

The performance of the students is a significant aspect in determining how well a college's instructors are doing. This report's objective is to suggest a new, clever method for predicting students' performance using support vector regression (SVR) that is improved by a dual algorithm (EDA). To the best of our knowledge, there aren't many papers published that have been established to forecast students' performance based on student behaviour, therefore the novelty of this investigation is to provide a fresh hybrid intelligent strategy in this area. The EDA-SVR scenario explicitly exceeded the other approaches, based on the data, by achieving smaller mean square error (MSE). In other words, EDA-SVR, with an MSE of 0.0091, performs better than DT, SVR, ANN, and PSO-SVR, which all have MSEs of 0.0315, 0.0242, 0.0232, and 0.0116, respectively. Other parameter estimation techniques, such as the direct method was evaluated, grid selection strategy, GA, FA, and PSO, are utilised in a comparison research to examine the effectiveness of EDA. The findings demonstrate that the EDA algorithm may successfully avoid local optima and blindness search, as well as accelerate resolution to the optimal method.