Loneliness Recognition Based on Mobile Phone Data

Nowadays, the definition of health is not only the absence of disease, but both physical and mental health. Loneliness as an important measure of mental health has become a topic that can not be ignored. In this paper, we study the problem about loneliness of individuals can be automatically recognized using mobile phone data (app usage data, call log, SMS, GPS data, Bluetooth proximity data and so on). In our study, we used 46 participants' data, divided them into risk and non-risk group based on self-reported scales for loneliness. We then compared the two groups to analyze the differences of mobile phone usage. And then we selected four kinds of classifiers - Support Vector Machine (SVM), Random Forest, Neural Network, and Gradient Tree Boosting (GTB) - to recognize loneliness automatically based on mobile phone data. The result showed that Random Forest can obtain the best performance with the accuracy of 70.68% for a 2-class loneliness recognition problem.