Desenvolvimento de uma ferramenta de learning analytics para o acompanhamento preciso de professores em atividades de programação

Abstract

Monitoring the learning process in programming activities is challenging, since traditional approaches tend to prioritize only the final result. Process mining techniques have been employed in this context, but many solutions rely exclusively on commits, which limits the granularity and temporal fidelity of analyses. This paper presents TKO-Analytics, a process-mining-based solution that uses granular telemetry collected by the TKO tool to assist instructors in monitoring the execution of programming activities. The solution integrates an Extract, Transform, and Load (ETL) pipeline for processing sub-commit events and code snapshots, applying process discovery and conformance checking techniques. Results are made available through a web dashboard that offers visualizations of student progress, quantitative conformance metrics, and detailed access to event logs and code evolution. The evaluation indicates that the use of continuous telemetry yields more representative process models and metrics that are more sensitive to student behavior, demonstrating the potential of the approach as a pedagogical support tool.