Machine Learning based Stream Selection of Secondary School Students in Bangladesh

Abstract

In the Bangladeshi education system, there are three stages up to the secondary school certificate (SSC)-the primary (Primary Education Completion Certificate, or PEC), middle school (Junior School Certificate, or JSC), and SSC. A separate stream has to be chosen after the eighth grade, which could be any of the following streams: Science, Business Studies, and Humanities. The selection of a stream is very important for their future higher studies and career planning. Usually, students take the decision of selecting a stream based on PSC and JSC results only. To address this challenge, we have collected a dataset from different Bangladeshi schools, which consists of PSC and JSC students' records. There are 26 data for each student record including subject-wise student results, parent's academic qualification, parent's profession, parent's monthly income, sibling information, district, etc. In the experimental analysis, a series of machine learning regression algorithms have been utilized. Moreover, we have employed various performance metrics in order to validate our model's performance. The experimental results demonstrate that among the regressors, extreme gradient boosting algorithm's performance were superior in both science and humanities streams. In the business stream however, Support Vector Machine's performance is considerably better. It is expected that the analysis will help prospective students and stakeholders in their future decisions. Moreover, we have utilized Local Interpretable Model Agnostic Explanations that helps to increase the interpretability of the model.

Attention to Monkeypox: An Interpretable Monkeypox Detection Technique Using Attention Mechanism

“…For this study, we have used two models for the interpretability of the decisions of the deep learning models. We have used LIME[52] and Grad-CAM[53]. LIME and Grad-CAM are prominent interpretability techniques that provide complementary insights into model predictions.…”