Detection of at-risk students with Learning Analytics Techniques

The way of teaching and learning in twenty-first century society continues to change. At present, a high percentage of teaching takes place through Learning Management Systems that apply Learning Analytics Techniques. The use of these tools, among other things, facilitates knowledge of student learning patterns and the detection of at-risk students. The aim of this study is to establish the most effective learning patterns of the students on the platform in a hierarchical order of importance. It was conducted over two academic years with 122 students of Health Sciences. The instruments used were the Moodle v.3.1 platform and the analysis of logs with Machine Learning regression techniques. The results indicated that the Automatic Linear Prediction Model detected by order of importance: average visits per day, student self-assessment questionnaires, and teacher feedback. The percentage variance of the final results explained by these variables was 50.8%. Likewise, the effectiveness of the behavioral pattern explained 64.1% of the variance in those results, finding three clusters of effectiveness in the behavioral patterns that were detected.