Discovering Unusual Periodically Human Activity Patterns Through Fuzzy High Utility Rare Itemsets Mining

Abstract

Monitoring the resident activities is a key goal to build a safe and convenient environment in a smart house system. One way to capture the activity is by installing motion sensors around the house. By collecting the triggered sensor state, we attempt to find unusual activity patterns to notify suspicious movements or health issues information as our main goals. This study focuses to express the nature of time interval of each activity more clearly using fuzzy set. Next, this study employs fuzzy high utility rare itemsets mining to obtain some odd activities with different quantity of time intervals. Hence, we can determine how peculiar the activity. In this study, we analyzed the public human activity recognition dataset based on each single resident and their activities. As the results, we successfully catch the uncommon activity compared to the daily ones either fishy movements or sick, the illogical movement order, and the false alarm due to the limitation of sensor transmission.