Abnormality Recognition Technology For Experimental Data Based On Deep Learning

With the continuous progress of big data, Internet of things, artificial intelligence and other technologies, we are in an era of information explosion. More and more test data are recorded. These data contain important information at all times. If we can find the relationship between these tests, we can have a deeper understanding and understanding of the test object. Based on the deep learning method, this paper studies the anomaly detection of multidimensional time series test data and the time series prediction of long test data, in order to effectively deal with these problems and improve the performance of the model. A long time series decomposition prediction model DMRFormer based on transformer is proposed. In order to solve the problem that it is difficult to find the dependencies between time series from complex and unknown time patterns, and the lack of prediction ability caused by transformer's self attention parallel structure, a progressive decomposition architecture based on transformer is designed and improved to improve the prediction performance. Based on the actual test scenario, using the relevant software development technology, a time series data management system in the test field is designed and implemented. The system mainly uses the proposed anomaly detection and prediction model to complete the prediction, anomaly detection and related governance tasks of time series data in the test scenario.