Distributed Decision Fusion for Large Scale IoT- Ecosystem

Abstract

IoT data analytics have numerous applications that generate huge data to gain new insights and information. How-ever, this work remains challenging due to the heterogeneity of IoT data sources, unnecessary data processing, uncertainty in decision-making, data biasness, and ever-increasing data size. To overcome these challenges, we propose distributed decision fusion framework for the large-scale IoT ecosystem. The proposed framework has divided into three-level. The first and second level provides the local decision of the small individual ecosystem using the filter method-based feature selection and dynamic classifier selection criteria for decision making; whereas the third level fuses the collected decision from the small ecosystems using Majority voting, Weighted majority voting and distributed Naive Bayes classifier. Lastly, we illustrate performance of the proposed solution on the US-Accidents dataset.