Descriptive Statistical Analysis and Discretization of Academic Data for Machine Learning Techniques

Every educational institute gathers voluminous data, which includes student enrolment data, student attendance, classroom participation, assessment results, course information, etc. Basic data analysis tools and techniques are not adequate to process and analyze a large volume of academic data. Machine learning or Data mining techniques have been applied by researchers which require pre-processing of the dataset. The discretization technique is commonly applied during data pre-processing to transform continuous feature values into discrete ones. The academic datasets are mostly continuous and hence require discretization. The paper attempts to apply descriptive statistical analysis to academic data to gain a better insight and understanding of the dataset. The observations obtained from descriptive analysis such as range, skewness, median, and mode are further applied in the discretization of the dataset. In the discretization process, both supervised and unsupervised methods have been used to convert continuous numeric values into discrete or nominal values. It has been found from the experiments that the CSForest algorithm has generated the best model for academic performance prediction. The proposed work has achieved a percentage accuracy of 97.10 % and outperformed the previous study.