Evaluation of machine learning techniques to classify code comprehension based on developers' EEG data

Abstract

Psychophysiological data such as brain waves have been used with machine learning techniques to classify the level of expertise and difficulty of software developers. However, little is known about the effectiveness of machine learning techniques (MLT) for classifying developers' code comprehension based on their brainwave data. This study evaluates the effectiveness of MLT's trained with EEG data to classify developers' code comprehension. Brainwave data collected from an EEG device while developers performed source code comprehension tasks was used to train the Neural Network, Support Vector Machine, Naive Bayes and Random Forrest classifiers. The effectiveness of these techniques was analyzed using accuracy, precision and recall. The Neural Network classifier, trained with EEG data and Principal Component Analysis, obtained 84% accuracy to classify code comprehension. Thus, the application of MLT to classify developers' code comprehension based on EEG data is possible.