Using Process Mining for Predicting Relationships of Couples Sitting on a Sofa

Abstract

This research is a synergy of Internet of Things (IoT), Process Mining (PM) and Behavior Analysis (BA) fields of study. In the first IoT-related part of the paper, a Wi-Fi P2P system (Peer to Peer) including a set of Smart Sofas, which were easily connected with a Xbox Kinect Camera without requiring a wireless access, was initially designed and developed. The Smart Sofas contained Weight Pressure Sensors, whereas, the Smart Cameras worked based on a Facial Recognition algorithm. The system was capable of identifying, recording and storing the exact location of a couple (i.e., two individuals) sitting on them in addition to their body language/gesture. Subsequently, 74 couples (or 148 persons) were voluntarily invited to join an experiment with the purpose of studying their behavior —within the duration of time they were used to getting back home from work— when sitting (or lying down) on a pair of smart sofas that was deliberately inserted in the living room of their home. The experiment lasted for one month and excluded the weekends so as to give the couples enough privacy they needed to carry on their normal life without any annoyance or disturbance. In the second part of the study, the Fuzzy Miner algorithm, which is a process discovery technique, was applied on the collected/synchronized (sofa and camera) data. To do this, the Disco Fluxicon, which is a Process Mining platform/tool, was used. The couples' data (i.e., the Sofa and Camera data) included 888 Events with a Mean Case Duration of 3.1 minutes. In the third part of the study, the data was divided into two separate event logs as the following: (1) The event log of the couples who showed/reported “Having Problems” in their relationship while taking part in the experiment, versus, (2) The event log of the couples who showed/reported “Not Having Any Problems”. Furthermore, the “Not Having Any Problems” data also was divided into another two sub-sets as follows: (i) Those who felt “Extremely Happy and Satisfied” versus (ii) Those who felt happy but in a very “Normal and Ordinary (Neutral)” way in their relationship. Subsequently, a statistical analysis of binary classification in terms of a Confusion Matrix and based on the F1-Score coefficient (or F-measure) was conducted so as to consider both the Precision and the Recall coefficients of the results. According to the findings of the study, with rather a high degree of accuracy (i.e., F-Score = 0.7563), it was realized that the couples who were “Not Having Any Problems” in their relationship showed tendency to represent/manifest one (or a mixture) of the following sitting positions while sitting on a sofa, respectively: “Legs on Lap”, “Cuddling in the Corner”, “Corner Cuddle with Tucked Legs”, “Side-by-Side (Touching but Not Cuddling)” and “Cuddling in the Middle”. And finally, the current work provides groundwork for further research and studies. In the future, more couples in a longer period of time (including the weekends) also will be analyzed and studied.