Exploring the Impact of Labeling on Psychophysiological Data Analysis

As the data grows, extracting information becomes more challenging. This problem is significant in domains where the ground truth is debatable and can be subjective, making the findings unreliable. Psychophysiological responses analysis is one of these domains, as the response to a stimulus can be heavily individual. We propose a simple yet robust approach to analyzing big psychophysiological data in a form of responses to external stimuli. Our approach uses a generic binary classifier and clustering techniques to generate a confusion matrix based on selected labels. These confusion matrices are analyzed to provide a detailed statistical summary and report the differences between models and responses within each group. We test our approach on a Emotion in Motion database that contains over 60,000 responses. The findings show that our binary classification approach provides insight and can be used to classify response based on the stimuli.