LSTM Learning With Bayesian and Gaussian Processing for Anomaly Detection in Industrial IoT

Abstract

The data generated by millions of sensors in Industrial Internet of Things (IIoT) is extremely dynamic, heterogeneous, and large scale. It poses great challenges on the real-time analysis and decision making for anomaly detection in IIoT. In this paper, we propose a LSTM-Gauss-NBayes method, which is a synergy of the long short-term memory neural network (LSTM-NN) and the Gaussian Bayes model for outlier detection in IIoT. In a nutshell, the LSTM-NN builds model on normal time series. It detects outliers by utilising the predictive error for the Gaussian Naive Bayes model. Our method exploits advantages of both LSTM and Gaussian Naive Bayes models, which not only has strong prediction capability of LSTM for future time point data, but also achieves an excellent classification performance of Gaussian Naive Bayes model through the predictive error. We evaluate our approaches on 3 real-life datasets that involve both long-term and short-term time-dependency. Empirical studies demonstrate that our proposed techniques outperform the bestknown competitors, which is a preferable choice for detecting anomalies.

Big-Data-Based Power Battery Recycling for New Energy Vehicles: Information Sharing Platform and Intelligent Transportation Optimization

“…This research has been well applied in the practicalengineering, and it has a very realistic reference value for government safety supervision, industry information tracing, national network planning and layout, enterprise investment planning, and recycling enterprise operation.Some researchers have achieved some research progress in using the state-of-the-art deep learning methods for anomaly detection or prediction. Wu et al [122]proposed a LSTM-Gauss-NBayes method, which was a synergy of the long short-term memory neural network (LSTM-NN) and the Gaussian Bayes model for outlier detection in Industrial Internet of Things. Xu et al [123]presented a novel unsupervised deep learning framework for anomalous event detection in complex video scenes,in which they utilized deep neural networks to automatically learn feature representations.This paper builds the battery recycling information platform by the big data, and by analyzing the operation mechanism of the platform and combining with the pipeline logic, it establishes the theoretical model of the power battery recycling platform based on big data, realizes the application model design of "information sharing, intelligent decision-making, optimization and integration" of retired power battery, and improves and enriches the application scope of big data.…”

Real-World Anomaly Detection by Using Digital Twin Systems and Weakly Supervised Learning

“…These networks are usually trained in unsupervised fashion with a dataset containing only data that reflects the normal state of operation. The reconstruction error is then used to distinguish anomalies with several different neural architectures, for example Autoencoders (AE) [14], Long Short Memory Network (LSTM) [6], Adversarial Autoencoder (AAE) [15] or Generative Adversarial Networks (GAN) [16]. When dealing with multivariate time-series data, Canizo et al [17] use a CNN-RNN capable to reach an AP score of 0.994 on a realworld dataset, but the proposed architecture is trained with a fully supervised approach.…”

“…Anomaly detection [3] refers to the problem of finding patterns in data that do not conform to expected or normal behavior. It is an active area of research with a wide range of application areas, such as energy [4], manufacturing [5], network sensors [6], health care and video surveillance [7]. Anomaly detection techniques based on machine learning can be separated into different types of approaches [8]: supervised approaches, where a sufficiently large set of training samples with labelled data is available; unsupervised approaches, where only the unlabelled measurement data is available; and weaklysupervised approaches, where a large amount of unlabelled data with a very small set of labeled data is available.…”

Anomaly detection‐based intelligent computing in internet of things and network applications

“…Such systems typically only use relatively a small amount of samples to train a goal‐oriented model which could be heavily affected by minor anomalies. Therefore, in the IoT and emerging network applications, it is urgent to detect anomalies from the collected samples to alleviate the side effect on model training 3,4 …”

“…Therefore, in the IoT and emerging network applications, it is urgent to detect anomalies from the collected samples to alleviate the side effect on model training. 3,4 Anomalies can be defined as the patterns that do not conform to expected behavior. 5,6 Actually, anomalies can appear in different forms in real applications, f.i., illegal intrusions on the Internet of services, 7 irregular behaviors in the scenario of smart city, 8 and abnormal events in smart agriculture.…”