Master's thesis submitted to the SIT (Schaffhausen Institute of Technology), Constructor Institute in partial fulfilment of the requirements for the degree of Master of Science in Computer Science and Software Engineering presented by Kyryll Puchkov under the supervision of Prof. Manuel Oriol co-supervised by Dr. Felix Fontein, Christian Vögeli.
When using learning platforms, many students seem to struggle without achieving success. The question arises: are they lacking effort or facing specific difficulties? This thesis aims to identify such students early on and enable teachers to respond promptly.
Using clustering techniques we identified two distinct clusters: one comprising students with higher effort but longer response times, possibly indicating learning disabilities, and another with lower engagement. The prediction models achieved a successful rate of 90% in accurately classifying and distinguishing between these clusters.
Furthermore, reaching a 90% accuracy score was achieved after 50 minutes of training time, equivalent to one week of real-time studies, allowing for accurate classification of children into specific clusters. These findings provide valuable insights for targeted interventions and optimisation of educational experiences for low-performing children.