Education data mining is data mining on an education dataset with an emphasis to understand the underlying patterns in student learning and its performance. The present study pertains to an analysis of performance of schools in Delhi, India using a decision tree approach. This helps to classify the schools that according to their performance. Data for government schools in Delhi, India for grade XII has been procured from the Delhi government education portal and the Unified District Information System for Education (UDISE) portal.
There are three measures of the school performance, namely, score in the individual subjects, quality index of a school and the pass percentage. Scores in a subject is the average score obtained by the students who have appeared in the given subject. Quality index of a student is the total score across the five subjects in which the student has appeared. Quality index of a school is the average value of quality index of the individual students of a school. Pass percentage is the percentage of students that have obtained the minimum passing score in all the subjects. The aim of schools is to maximize their pass percentage (typically, 100%) as well as the scores, however, first it is desired to improve the pass percentage to have minimum levels of attainment for all the students.
The school input data used in the analysis pertains to its infrastructure that includes classrooms, playground, library, etc; its operations that comprises the medium of instruction, implementation of continuous and comprehensive evaluation, number of working days, working hours per day, number of inspections by external officials and other concerned authorities; school enrolment, and number and distribution of teachers; type of school (boys’, girls’ or co-educational) and school category (based on grades offered). The variables that are correlated have been removed to yield a reliable rule set.
The schools are classified based on whether they achieve the minimum acceptable level of performance. The schools are classified with two target labels: (i) those that achieve minimum acceptable performance, (ii) those that do not have minimum acceptable performance. The decision rule set can be used by the policy makers to identify and take appropriate corrective actions in the schools to help them with minimum acceptable performance. The work can be extended to other classification algorithms as well.