The PASSt Project: Predictive Analytics and Simulation of Studies aimed at Quality Management and Curriculum Planning

Abstract

Quality management has become a crucial factor for improving student success, with reporting being widely used to scrutinize curricula for possible bottlenecks and resource deficiencies. Predictive capabilities in that context have, however, been often limited to simple regression models acting on historical data, which might not always be available when curricula change often; furthermore, work in curricular planning often demands “what if”scenarios that are beyond extrapolation, such as determining the influence of changes in procedure on student success, which in itself is based on a multitude of intertwined factors such as social background and individual performance. In the PASSt project, we have been using Machine Learning and Agent-Based Simulation for Predictive Analytics in that sense. As a result, we have been developing an extensive toolset for curriculum planning which we want to outline in this paper, together with some lessons learned in that process. Our work will help practitioners in higher education quality management implement similar methods at their institutions, with all said benefits.