Application of machine learning algorithms for detecting anomalies in big data streams of sensor readings

Abstract

An overview of the basic anomaly detection algorithms and their implementation in the applied problem for big data streams processing of sensor readings is considered. New approaches for detecting anomalies based on various architectures of neural network autoencoders are proposed and their comparative analysis is considered. Keywords anomaly detection, machine learning, neural networks, recurrent neural network, encoder, decoder, fully connected layer, latent vector