Student Progression System using Descriptive and Predictive Analytics

Abstract

Analytics is a practice of dividing a problem into simpler parts and then use their applications to make decisions better. It is only a way of thinking not a tool or technology. It has diverse applications across the globe as education, retail, marketing, gaming and health care. Data analytics play an important role in any organization, based on relevant facts that will allows making a better decision. Progression of students greatly affects the educational organization’s future. Analysis of the academic dataset could reveal important insights, which if properly used can help students for their progression. In this work, the Descriptive and Predictive Analytics has been applied on the Academic dataset of students. Descriptive Analytics is the summarization of the past data and generates some useful patterns from that data. This work focuses on analyzing and querying large academic dataset for generating Student Progression using visualization and dashboards. Presently projects on Progression Systems exist but no descriptive or predictive analytics has been performed on these datasets. The proposed system collects data from different departments of University, store data into the large data warehouse of the University and generate validated set of Key Performance Indicators (KPIs), based on the past dataset of student Academic. These KPIs have been obtained after applying Statistical techniques on various sets of dimension on the academic datasets. After completion of this step, the focus was on predicting the performance of the student’s by using cluster analysis approach. For this work, the cube designing has been done and three clusters have been fabricated by using k-means clustering algorithm. These clusters placed the similar attribute in one class and the objects with different attributes in other class, on which various models have been applied for predicting the performance of the students. The deep learning model has given the highest accuracy that is 99.02% in comparison to all other models such as Naïve Bayes, Decision Tree and Linear Discriminant Analysis (LDA).