EMOTION ANALYTICS WITH PROCESS MINING

Abstract

This research builds on the intersection of ‘process mining’ and ‘emotion analytics’ in order to discover and investigate the emotional patterns of students during the StudentLife project based on data collected by smartphones through PAM application (i.e., Photographic Affection Meter). The main objective of the study is to analyze and predict the relationships (or continuity) amongst 16 common emotional indicators based on the ‘Russel’s Circumplex Affect Grid’ and by means of Fuzzy Miner (supported by Disco Fluxicon) and Dotted Chart Analysis (supported by ProM) process mining tools and techniques. Accordingly, the current work is divided into two main parts. In the first part, a pre-processing (or data preparation/cleansing) approach via Python programming was done in order to change the format of the initially collected event logs from JSON to the appropriate format/structure. In the second part, the emotional datasets were analyzed using the above-mentioned techniques. To do this, new groups/clusters of contexts were categorized and pre-defined. The third part of the study deals with data interpretation and discussion of the obtained findings. The proposed/applied approach was capable of providing “frequency-based” models/graphs of the students’ behavior, both before and after the experiment, in terms of 5 categories: “Minimal, Minor, Moderate, Moderately Severe, and Severe” depression-related emotional trends ranging from Low Severity (NA) to High Severity (PA). This research provides groundwork for further and future studies.