Performance Prediction of Learning Programming – Machine Learning Approach

Abstract

Teaching and learning programming is a challenge faced by many educational institutions. In this paper we described using machine learning with data mining techniques to predict the performance of students using SMAC-based (social, mobile, analytics and cloud) programming learning tool (SPLT) that we developed for students to learn computer programming. Being able to predict and know students’ performance has many advantages from educators’, students’ and administrators’ perspectives for better learning, teaching, pedagogy design and institutional management. With the designed SPLT, experiments were conducted with 71 students from higher institutions who are learning computing programming. Various data were collected during the course of the experiment such as participants’ demographics, programming background, logged data in SPLT, chat logs, pre- and post-tests scores for data mining attributes. Four classification algorithms were used to develop the classification model for the two datasets we prepared. WEKA was used to clean raw data, train and measure the performance of the models. Our experiment indicated that we were able to predict students’ performance consistently with high accuracy of F-score using Random Forest classifier.