Identifying Student Learning Patterns with Semi-Supervised Machine Learning Models

Abstract

One of the benefits of adaptive learning systems is that they allow students to work at their own pace. Because of this, students may exhibit drastically different learning patterns, some of which are symptomatic of misuse or suboptimal use of the system, or simply of possible inadequacy in the system. Identifying such patterns allows the system or the instructor to take corrective action to ensure that students are having a successful learning experience. ALEKS, which stands for “Assessment and LEarning in Knowledge Spaces”, is a web-based artificially intelligent learning and assessment system. In this work we attempt to identify and classify various learning patterns that students exhibit while working in the ALEKS learning mode. To do this, we first build a set of statistical features for describing the learning behaviors that students exhibit. After using these features to identify an example set of students, we use semi-supervised machine learning techniques combined with an artificial neural network to apply these classifications to the rest of our dataset.