From error patterns to learning opportunities: ML classification and GenAI synthesis in coding education

Abstract

Coding education is increasingly turning to technology to provide students with personalised and effective learning experiences. This paper presents a novel approach that leverages machine learning (ML) classification and generative AI (GenAI) synthesis to identify and replicate common coding mistakes, turning them into valuable learning opportunities. The ML model is trained to classify erroneous code samples into predefined categories, providing insights into the types of mistakes students frequently make. Simultaneously, the GenAI model generates synthetic yet realistic erroneous code based on these patterns, allowing students to practice identifying and correcting these mistakes. By incorporating both ML classification and GenAI synthesis, this system enhances coding education by fostering targeted practice, improving error identification skills, and providing a deeper understanding of common coding pitfalls. This report explores the technical implementation, educational benefits, and potential for broader adoption of this integrated approach in coding education.