Scalable Machine Learning Framework for Efficient and Accurate Prediction and Classification in Smart Environments

Abstract

Technologies such as the Internet of Things (IoT) and edge computing have led to rapid development in the field of smart systems. As a result, the emergence of artificial intelligence models that are scalable, accurate, and consume less amount of resources has been witnessed. In order for the traditional and centralized learning techniques to function, they require a significant amount of data collection. This results in high communication overhead, increased latency, and unnecessary usage of resources. Due to this reason, these solutions fail to perform well in automated systems that function in real-time. The current research proposes an architecture for machine learning that is scalable and is aware of the resources. It is designed to predict and classify accurately in intelligent systems that are heterogeneous. The framework makes use of distributed learning techniques. It also employs adaptive resource management to increase the efficiency of computing, communication, and energy consumption without degrading the performance of the model. The technique works efficiently in both.