A Fundamental Study of Social Confusion Detection System by Machine Learning using Human Flow Data

Abstract

Recently, people have become more aware of disaster prevention due to the effects of natural disasters that cause serious damage. The concept of disaster prevention consists of three elements: "pre-disaster prevention" before disaster, "prevention of damage expansion" and "disaster recovery and reconstruction" after disaster. The current disaster prevention system can observe before the disaster; however, cannot observe the damage situation after the disaster. Delaying an initial response to a disaster has a great impact on rescue activities and secondary disasters, so a disaster prevention system after disaster is very important for disaster prevention. In recent years, anomaly detection systems using big data have been attracting attention as disaster prevention systems after disasters. In this thesis, we focused on people flow data among big data, and discussed a social confusion detection system by machine learning using human flow data in normal and abnormal times.