Predictive Analytics for Student Counseling and Academic Capacity Planning: A Case-based Decision Support System

Presentation Summary

Bodo Rieger, Daniel Poppelmann and Sonja Schulze (University of Osnabrueck)

We present an application for the rolling prediction of future course-enrolments applying a refined case-based reasoning (CBR) approach. As a first step towards a sophisticated predictive analytics tool this project aims at supporting executives, faculty and students in their course- and resource-planning processes. Using the open-source jColibri framework we designed a case-base that consists of more than 900 heterogeneous student cases modeled in an object-oriented manner, including personal attributes (e.g. age, sex, a-levels grade) and course histories (including current GPA, failed and completed courses, etc.). The retrieve-phase of the CBR-cycle introduces a dynamic case interpretation with regards to the stored cases’ descriptions and solutions. A rule-based reasoning approach is integrated in the revise-phase to verify that proposed solutions (i.e. a student’s predicted future study plan) meet the examination regulations. Next, solutions are presented to the respective students who benefit from a proposal of selectable future courses and are able to validate/change the proposition against their preferences. In order to support executives and faculty revised solutions are stored within a database and processed for multidimensional/analytical and standard reporting on different levels of aggregation (e.g. predicted number of enrolments for class Accounting I in Fall 2012).