Analysis of Student Behavioural Patterns by Machine Learning

Abstract

AbstractAn important task in education field is discovering student behavioural patterns to take timely action to improve student activities or grades. Sometime students may fell into depression due to misunderstanding of subjects or due to low grade which leads into abnormal behaviour, and by identifying such abnormal behaviour, institutions can take necessary steps to improve student’s condition. For this research, questionnaire method is used which includes collecting student data through survey and analyse students’ behavioural patterns. However, results by this method are not effective or accurate as this method largely relies on feedback data. So to solve this problem, an unsupervised clustering approach can be used. This produces relatively accurate results. The proposed framework integrates two unsupervised clustering approaches, i.e. density-based spatial clustering of applications with noise (DBSCAN) and k-means. The students data is collected from Kaggle data sets. The proposed framework extracts necessary behaviour features by statistics and entropy to find both anomalous behavioural patterns and main stream patterns. To predict whether the student is low active or high active or medium active, we can use supervised techniques as unsupervised clustering approaches are meant to form clusters. These findings can help students to improve their grades and personality and organization can also take appropriate steps to help students by providing better services and administrations such as psychological consultations and academic advices.KeywordsEnsemble clusteringDB SCANK-meansSupervised techniques