Research On The Framework Of Mining On Multiple Time Series

Time series is a set of time-series data set, which exists in a wide range of commercial, transport, industry and other industries. Analysis of time series data can reveal the internal rules of things in movement, change and development, and have important practical significance for the people to understand things correctly and make a scientific decision.In this paper, the specific research background is control for hot-rolled. In the rolling process, many anomalies are caused by a number of reasons, and there is existence of the relationship between the reasons. For this, the paper presented a multiple time series framework of abnormal pattern mining. First of all, the framework uses method of pattern representation to compress the data, and make abnormity detection for the compressed data. And then mining on the basis of abnormal pattern, in order to found some Decisive and guiding informations for the decision-making of companies.Pattern representation of the framework, through the basic idea of the edge operator in the field of digital image, combinating the edge operator and the characteristics of time series, there is time-series piecewise linear representation.Anomaly detection of the framework, adding the idea of a sliding window based on TOD anomaly pattern discovery algorithm, for detection of local anomaly.In the mining sequential patterns, the framework uses an improved algorithm, which is based on PrefixSpan. By removing the non-frequent itemsets and determining the relationship between projection sequence number and the number of minimum support, to reduce unnecessary storage space and improve the query speed. Mining of multiple time series of the framework used frequent pattern discovery algorithm in non-synchronous multi-time series. Experiments for the above-mentioned Algorithm demonstrated the feasibility and effectiveness of every algorithms.Experiments show that this framework has been completed well in abnormal patterns detection of the process of steel rolling, as well as the mining of the interrelationship between various factors. The framework can play a good role for improving the quality of steel products for the enterprise and controlling the quantity of steel production.