An IoT based Machine Learning Technique for Efficient Online Load Forecasting

Abstract

Internet of Things (IoT) networks are computer networks that have an extreme issue with IT security and an issue with the monitoring of computer threats in specific. The paper proposes a combination of machine learning methods and parallel data analysis to address this challenge. The architecture and a new approach to the combination of the key classifiers intended for IoT network attacks are being developed. The issue classification statement is created in which the consistency ratio to training time is the integral measure of effectiveness. To improve the preparation and assessment pace, it is suggested to use the data processing and multi-threaded mode offered by Spark. In comparison, a preprocessing data set approach is proposed, resulting in a significant reduction in the length of the sample. An experimental review of the proposed approach reveals that the precision of IoT network attack detection is 100%, and the processing speed of the data collection increases with the number of parallel threads.