A Smart Data Pre-Processing Approach to Effective Management of Big Health Data in IoT Edge

Abstract

Background: An IoT Big Data analysis platform should be able to dynamically manage IoT data and it should be appropriate the fundamental components, known as the 5V's of big data. Therefore, speed and accuracy are two important criteria to consider. In this context, there are no similar studies that prioritize speed and accuracy criteria in big health data. It is thought that this study and the experimental results obtained are a new approach in the field of healthcare, hence it will add novelty to the studies to be carried out. The main objective of this paper is to detect anomalies at the edge of IoT for the effective management of big health data. Methods: This study focuses on detecting anomalies on the data stream created with IoT sensors between the sensing and network layer. The classification success and data processing speed of the random cut forest, logistic regression, Naive Bayes, and neural network algorithms used for anomaly detection are compared. In order to detect anomalies in a data stream consisting of temperature, age, gender, weight, height, and time data and compare algorithms. Results: The speed and accuracy performances of ML Algorithms were compared. The performance comparison shows that the LR algorithm will be more successful in IoT systems in terms of speed, although it is very close to the RCF in terms of accuracy. Conclusion:The experimental results show that using ML algorithms on IoT edges will help make effective and timely decisions in the healthcare domain. Thus, the big data generated by the IoT sensing layer in healthcare will be formed at a more manageable level. Also, thanks to this, service providers, users, and other interested sides will be minimally affected by the negative effects of anomalous data.