A Framework for Selection of the Most Suitable IoT Device based on the Execution Context

In modern era, all the digital devices are interconnected using the IoT ecosystem. Based on external trigger event or user's command, these devices work in sync and provide response to the user. These actions on the IoT devices can be of two type, actions which can be performed by a unique device, such as washing machine, microwave etc. or actions which can be performed by more than one device, such as music playing, weather queries, Question and Answer chats etc. In a domestic IoT environment, more than one IoT device can perform same task. Our paper proposes a novel framework to select the most suitable device for performing the task, based on execution context. We explain the data collection and data generation method using a web portal for the IoT digital devices. For increasing the dataset, to cover all possible device states, we have explained the approach with code flow. Then we have explained the machine learning models to cover two IoT use cases in domestic environment covering daily use scenario. Decision tree and random forest based algorithms are proven to perform good with accuracy above 85% on our generated dataset. Aim of the paper is to provide an end to end framework, by which any researcher or developer, can define machine learning based policy to execute the tasks efficiently in multi device IoT environment.