Investigating prediction modelling of academic performance for students in rural schools in Kenya

Abstract

Academic performance prediction modelling provides an opportunity for learners' probable outcomes to be known early, before they sit for final examinations. This would be particularly useful for education stakeholders to initiate intervention measures to help students who require high intervention to pass final examinations. However, limitations of infrastructure in rural areas of developing countries, such as lack of or unstable electricity and Internet, impede the use of PCs. This study proposed that an academic performance prediction model could include a mobile phone interface specifically designed based on users' needs. The proposed mobile academic performance prediction system (MAPPS) could tackle the problem of underperformance and spur development in the rural areas. A six-step Cross-Industry Standard Process for Data Mining (CRISP-DM) theoretical framework was used to support the design of MAPPS. Experiments were conducted using two datasets collected in Kenya. One dataset had 2426 records of student data having 22 features, collected from 54 rural primary schools. The second dataset had 1105 student records with 19 features, collected from 11 peri-urban primary schools. Evaluation was conducted to investigate: (i) which is the best classifier model among the six common classifiers selected for the type of data used in this study; (ii) what is the optimal subset of features from the total number of features for both rural and peri-urban datasets; and (iii) what is the predictive performance of the Mobile Academic Performance Prediction System in classifying the high intervention class. It was found that the system achieved an F-Measure rate of nearly 80% in determining the students who need high intervention two years before the final examination. It was also found that the system was useful and usable in rural environments; the accuracy of prediction was good enough to motivate stakeholders to initiate strategic intervention measures. This study provides experimental evidence that Educational Data Mining (EDM) techniques can be used in the developing world by exploiting the ubiquitous mobile technology for student academic performance prediction.