Detecting abnormal human behavior by using data from user's portable devices

Abstract

A method for detecting abnormal behavior involves constructing words and text documents based on data acquired from mobile phone sensors during defined time intervals. The time intervals are defined based on data from mobile phone sensors indicative of usage patterns of the mobile phone user. Words are constructed for each time interval as a vector including the time interval and sensor-based feature levels. Each sensor-based feature level is mapped to a range of values of a sensor-based feature that are extracted from the sensor data. The text document is constructed from the words based on the time intervals and the sensor-based feature levels. A current routine for each time interval is determined using topic modeling based on the words that most frequently appear in the text document. An alert is generated if the current routine for any time interval deviates from a past routine for a corresponding past time interval.