Time Series Data Prediction And Correlation Mining Model Based On Recycle Evolution Networks

Previous studies on time series data mainly focus on the predictive analysis,cluster analysis and association analysis.The existing research methods mainly have the following disadvantages: firstly,most of the existing time series models use a single network model to analysis data,ignoring that different data have different characteristics.As a result,using single network model to predict,the accuracy is not high.Secondly,in the existing clustering algorithms,most of them using mathematical distances of the data and different rules for clustering,which results in that clustering cannot find the implicit relation existing naturally between data.Thirdly,in the existing correlation mining,mathematical methods are usually used to calculate the correlation between data and build a complex network for mining and analysis.Implicit relationships within data are ignored,resulting in a relatively single perspective of data analysis.To solve the above problems,classical mathematical relationships were used to mine data based on relational mining in previous work,this paper mainly proposes recycle evolution networks prediction and association mining algorithm framework.This framework builds complex networks for data mining and knowledge discovery,which based on machine learning to analyze associations between data.At the same time,the framework,which is based on recycle evolution networks,predicts and clusters the data,and uses complex networks as the basis for data association mining.With the financial stock market and ocean big data as the background,and the plate stock price and marine hydrodynamics elements as the object,the main research contents are as follows:Firstly,to solve the problem that single model cannot predict accurately,this paper proposes recycle evolution networks model.The model is composed of multiple single models and each model can predict and analysis different kinds of data to improve the accuracy of prediction.The model is a chain structure composed of single models.Each single model contains a numerical prediction unit and an error prediction unit.The error prediction unit is used to predict the data error predicted by the numerical unit,and then classify the data according to the data division rules.Finally,train the data again and finally stop data training through the termination mechanism.In terms of application,the model predicts ocean surface temperature data and compares with other models to verify that the model can improve the accuracy of data prediction and is superior to other models.Secondly,to solve the problem that the existing clustering methods are based on the data distance clustering method,this paper proposes two clustering methods which are based on different strategies recycle evolution networks,the data clustering method based on the mean of the error and the data clustering method based on the experience of the experts.These two clustering methods are based on cyclic evolution network prediction and take data partitioning as the core.Finally,the model is applied to the single-point ocean surface temperature data to classify.It can be seen that the clustering data has spatial continuity and seasonal periodicity.Therefore,this clustering method is meaningful.Thirdly,aiming at the problem of constructing complex networks with existing mathematical methods for association analysis,this paper proposes to use machine learning to establish data relationships and construct complex networks.The model calculates the coincidence degree of the data by using the clustering results of the recycle evolution networks.Then it constructs a complex network based on the coincidence degree of the data.In addition,it analyzes the average weighted strength of the nodes,the average path length,the degree of modularity of the complex network,and the feature vector center performs data association mining.After using the mathematical relationships in the previous work to build a complex network and analyze financial data,it can find out the plate linkage features and plate drift of stock data.In the process of study multi-point data of ocean dynamics elements by using complex network which is based on the relationship of machine learning,it can discover the remote correlation characteristics between the data.In this paper,the time series prediction and correlation mining framework based on recycle evolution network is an integration model of prediction analysis and mining analysis based on data-driven multi-network model,data partition rules as the core,and complex network analysis as the basis.