Multiple Time Series Association Rules Based On Change Characteristic Discretization

With the large-scale deployment of different types of sensors,such as temperature sensors,pressure sensors,humidity sensors on construction machinery,these sensors generate a large number of multi-dimensional time series data.Through the time series data association analysis of these sensors,it can be used to discover the potential relationship of working conditions of various parts of construction machinery.This potential relationship will provide support for the early warning and analysis of faults in the construction machinery system.However,the traditional Apriori correlation algorithm can not be used directly because of the high dimensionality and large data volume of time series of construction machinery.Therefore,based on the real time data of construction machinery,this paper presents a method of relevance analysis for the multi-dimensional time series of construction machinery in view of the specific characteristics of time series of construction machinery.Firstly,the whole framework of the multi-dimensional time series correlation analysis system is designed.The framework is divided into pre-processing module,discretization module and association rule extraction module.The preprocessing module is responsible for data cleaning and normalization,normalizing the range of values for each time series,and making each time series consistent with the Gaussian distribution.Then,the discretization module expresses and discretizes the normalized time series,that is,transforms the numerical data into character data.Finally,the association rule extraction module extracts the effective association rules from the discretized time series by using the improved time characteristic association method.This article then describes in detail the techniques used by the discretization module.In this paper,the PAA + SAX discretization method is evaluated and it is found that the discretization method of PAA + SAX has some shortcomings.First,the PAA representation changes the time series normalization results.And the PAA representation ignores the key information of the original time series.Then SAX discretization method can only represent the size of time series characteristics,and SAX on the premise that the time series with Gaussian distribution.However,most of the data set in this paper is not Gaussian.Therefore,this paper improves the SAX discretization method so that it can also deal with non-Gaussian time series.In this paper,we propose a discrete discretization method,which can process time series of non-Gaussian distributions and preserve the key information of the original time series.Finally,this paper describes the temporal association method used by the association rule extraction module.This method improves the traditional Apriori algorithm.Because traditional Apriori algorithm does not consider the time characteristics of items,it is necessary to improve the generation algorithm of candidate items,support method and association rule generation algorithm.So that the improved time characteristic correlation method can not only deal with single time series,but also can deal with multiple time series.In addition,the frequency threshold and the support threshold are used to obtain the anomaly pattern.Finally,the key time segment correlation method is proposed for the periodicity of the data set in this paper,and the running time of the correlation algorithm can be reduced.