Extração descentralizada de conhecimento associativo para internet das coisas

Abstract

Identifying user behavior patterns is one of the features that can be incorporated into the Internet of Things. Finding standards and using them as knowledge for decision making can provide ease, comfort, practicality and autonomy for the execution of daily activities. Although knowledge extraction is common in centralized intelligent environments, its execution in a decentralized architecture is still a relevant computational challenge onsidering the storage and processing constraints of IoT devices. This dissertation describes a method for mining implicit correlations between IoT device action patterns through embedded associative analysis. Based on the metrics support, confidence and lift, the method identifies the most relevant correlations between a pair of actions from different devices and suggests to the user the integration between them through HTTP requests. Experimental results show that, on average, the most relevant rules for both architectures are the same in 99.75\% of cases. In addition, the proposed method identified relevant correlations that were not identified by the centralized architecture. This research emphasizes that device action pattern analysis is an efficient approach to provide a highly integrated and intelligent environment by circumventing single point failure problems and excessive data storage on IoT devices.