Analysis of the Socio-economic Factors of Poor Academic Results and Predicting Probable Solutions of Major Factors

Abstract

Measuring student performance based on both qualitative and quantitative factors is essential because many undergraduate students could not be able to complete their degree in recent pasts. At present, students’ dropout rate in university is gradually increasing and many bright students sometimes just cannot cope with the universities. This research is mainly based on finding the reasons for students’ different types of results and then predicting students’ performance based on those significant factors. Varied social and economic theories have been used for the identification of several factors for the surveys and stratified random sampling technique has been used for the collection of data. Significant factors were later identified using the analysis of variance (ANOVA) test. Then, two popular supervised machine learning algorithms have been used for classifying students’ different levels of results and predicting students’ performances, these are support vector machines (SVM) and random forests (RF) which are tremendously used in classification and regression analysis. The input dataset for both training and testing were taken by merging the values obtained from two surveys done on students and experts using adaptive neuro-fuzzy inference system (ANFIS). The result exhibits that RF can perform the classification of multiple classes based on many distinguishing features with more confidence than SVM. Afterwards, important factors responsible for students’ poor performances were examined to find the probable solutions in contrast to those. This proposed model can also be applied to predict course-wise students’ performances.

Keywords: SVM, RF, Performance Prediction, Socio-economic factors, Machine Learning