Anomaly detection model using time serise dataset of small manufacturing industry

Abstract

As artificial intelligence technique is generalized widely used in industry area. so, there are attempts to anomaly detect by using deep learning in small manufacturing industries. However, it is difficult for small manufacturing industries to have an artificial intelligence infrastructure. The nation support data set of open small manufacturing industries for solve these problems and help. This paper proposes an anomaly detection model for time series data using this data set. The propose LSTM-SVDD anomaly detection model is that combines the LSTM model widely used in time series data with the SVDD model widely used in anomaly detection. The propose model is that learns the range of normal data and detects data out of this range as abnormal data. It is confirmed that the data distribution of the test data not used for learning predicted similarly with prediction results. A performance indicator ROC is also high at 96.31. the proposed automatic anomaly classification model is expected that can be used in small manufacturing industries field that are limited in the construction of artificial intelligence infrastructure.