Integrating Decision Trees with Metaheuristic Search Optimization Algorithm for a Student’s Performance Prediction

The main objective of educational institutions these days is to refine their student's performance in the overall grind. One way to achieve the zenith for these is to identify factors including curricular and extracurricular activities of students. The basic aim of the proposed work is to predict a learner's performance based on certain parameters such as the number of absences, family background, eating and sleeping habits, etc. This kind of performance prediction can be used by facilitators or faculties to help learners improve their academic conduct. In this paper, the Classification and Regression Tress(CART) algorithm is used for the construction of the initial population of decision trees. The dataset is pursued in two ways for construction of decision tree: 1) It is divided into five different sub-datasets depending on the similarities of attributes and 2) Original one single dataset is considered. To make an optimal prediction, a metaheuristic optimization algorithm that is, a genetic algorithm has been used. According to genetic algorithms, the best chromosome(solution) to the problem survives amongst all possible chromosomes(solutions). However, its performance might get deeply affected by the choice of the genetic operators used in various steps of the genetic algorithm. There are four different classes namely Class 1, Class 2, Class 3, Class 4 based on the different attribute values of the instances. The proposed approach yields 91% accuracy for the final prediction of classes, which is quite encouraging as compared to the other existing works in the same domain.