A Data-Driven Causal Modelling Analysis of Socio-Economic Factors and Its Impact on Student’s Performance: A Case Study of a Junior High School in Bali

Abstract

Efforts to understand student performance have long been a highly-researched topic in the field of applied education computing. Current research in the field still places its focus on understanding and analyzing student performance using definitive variables such as the student’s scores and their cognitive capabilities, which by themselves already explain the student’s performance. The great diversity of Indonesian culture, which includes people from a wide range of socioeconomic origins, makes it all the more surprising that so little research has been done to examine the hidden socioeconomic aspects that may affect student performance. Research conducted on a single school may not be generalizable because of the diversity among them in terms of the elements that affect students' academic outcomes. In this investigation, we employ a causal modelling strategy that is data-driven to examine academic achievement. Data was retrieved from a public junior high school in Bali, Indonesia, and then processed with the Non-combinatorial Optimization via Trace Exponential and Augmented lagRangian for Structure Learning (NOTEARS) and Bayesian Network algorithm in order to discover a latent causal structure and the effect between variables discovered from the structure. Findings show that the average skill score of a student is significantly influenced by the distance from school, the education level and income of parents, and their place in the family. Meanwhile, the average knowledge score is mainly influenced by the average skill score, the order in the family, and the parent’s income level. The results of the study also show potential for practical implications where schools, researchers, and governments, can rethink their approach to education by analyzing data with the proposed approach. The limitations of this study include the quality of data to discover patterns and the limited number of variables used to study student performance factors. Future research may consider the use of more holistic, complete variables in order to discover more insights regarding student performance.