Self-evolving intelligent algorithms for facilitating data interoperability in IoT environments

Internet of Things (IoT) is predicted to connect 20.4 billion devices in 2020 and surge to 75 billion by 2025. Such a connected world where machines will communicate with other machines opens up huge opportunities and a very different way of life, with smart homes, self-driving vehicles and wearable devices. It is expected that such interconnectedness will enable the capture of events as data in real time and provide actionable insights to people and organizations to maximize efficiencies, be pro-active and more effective. Interconnected devices will require interoperability, and the seamless, secure and controlled exchange of data between devices and applications has been called data interoperability. Such a dynamic and volatile environment with a wide diversity of data will require a new breed of intelligent algorithms with the ability to adapt and self-learn as well as envisage and analyse events at multiple levels of abstraction to gauge association and interrelationships. This research proposes three algorithmic requirements for intelligent algorithms in such IoT environments: unsupervised self-learning capability, ability to self-generate to the environment and incrementally learn with temporal changes. The paper first presents empirical results with real data from a fire department in Australia to highlight the need and value of IoT and data interoperability. Dynamic Self Organizing Map based unsupervised algorithms which satisfy the requirements are described and further empirical results are presented to validate the required functionality of these algorithms.