Open Access

ARTICLE

Evolutionary Algorithm Based Feature Subset Selection for Students Academic Performance Analysis

by Ierin Babu^1,*, R. MathuSoothana², S. Kumar²

1 Department of Computer Science and Engineering, Noorul Islam Centre for Higher Education, Kanyakumari, 629180, India
2 Department of Information Technology, Noorul Islam Centre for Higher Education, Kanyakumari, 629180, India

* Corresponding Author: Ierin Babu. Email:

Intelligent Automation & Soft Computing 2023, 36(3), 3621-3636. https://doi.org/10.32604/iasc.2023.033791

Received 27 June 2022; Accepted 23 November 2022; Issue published 15 March 2023

Abstract

Educational Data Mining (EDM) is an emergent discipline that concentrates on the design of self-learning and adaptive approaches. Higher education institutions have started to utilize analytical tools to improve students’ grades and retention. Prediction of students’ performance is a difficult process owing to the massive quantity of educational data. Therefore, Artificial Intelligence (AI) techniques can be used for educational data mining in a big data environment. At the same time, in EDM, the feature selection process becomes necessary in creation of feature subsets. Since the feature selection performance affects the predictive performance of any model, it is important to elaborately investigate the outcome of students’ performance model related to the feature selection techniques. With this motivation, this paper presents a new Metaheuristic Optimization-based Feature Subset Selection with an Optimal Deep Learning model (MOFSS-ODL) for predicting students’ performance. In addition, the proposed model uses an isolation forest-based outlier detection approach to eliminate the existence of outliers. Besides, the Chaotic Monarch Butterfly Optimization Algorithm (CBOA) is used for the selection of highly related features with low complexity and high performance. Then, a sailfish optimizer with stacked sparse autoencoder (SFO-SSAE) approach is utilized for the classification of educational data. The MOFSS-ODL model is tested against a benchmark student’s performance data set from the UCI repository. A wide-ranging simulation analysis portrayed the improved predictive performance of the MOFSS-ODL technique over recent approaches in terms of different measures. Compared to other methods, experimental results prove that the proposed (MOFSS-ODL) classification model does a great job of predicting students’ academic progress, with an accuracy of 96.49%.

---

Keywords

---