Authors :

Kumar Mohan, Mohamed Kuresh Safir, and Rajesh Natarajan

Address :

University of Technology and Applied Sciences, Sultanate of Oman

University of Technology and Applied Sciences, Sultanate of Oman

University of Technology and Applied Sciences, Sultanate of Oman

Abstract :

Wearable sensors (WS), depending on where they are placed and what they are used for, transmit data at regular or irregular intervals. Accurate and consistent analytics require handling each asynchronous data stream separately. This provides dependable responses and solutions for diagnosis and prophylactic measures. In order to handle asynchronous data streams, a Variant Data Analytical Model (VDAM) is presented in this study. In this theory, data analysis takes into account changes in the size, sources, and priority of the data stream. For priority-based variation analysis, a standard prediction classifier is employed to separate variants from consistent data streams. The probability instances of the classifier are trained using these segregated and non-segregated instances. As a result, the classifier produces balanced, parallel, latency-controlled analytics of the input stream. Additionally, the analytical model may be easily adjusted to accommodate data streams derived from WS sources, thereby decreasing pipelined processing errors. Based on the classifier output, a response or set of recommendations for the diagnosis is given, limiting the high processing rate. p>

Keywords :

Data Analytics, Parallel Processing, Prediction Classifier, Wearable Sensors.