Research On Time Series Prediction Based On Optimal Weighted Combination Model

With the rapid development of a new generation of digital technology with the core of"cloud computing,big data,internet of things,mobile internet,artificial intelligence and blockchain",data mining technology is gradually being applied to various fields.Among them,time series prediction,as an important branch of data mining technology,has profound significance and important research value in many fields because it can predict the changes of data in a future period of time by analyzing the internal laws of data.However,with the continuous expansion of data scale and the continuous improvement of complexity,the demand for time series prediction accuracy in various fields is also gradually improving.A single prediction model has gradually been difficult to meet the needs of practical engineering application and fine planning and decision-making in various fields under the background of big data.In order to improve the prediction accuracy of existing prediction models in different fields and scenarios,combined with three kinds of combined prediction methods:parameter optimization,component integration and data preprocessing,this paper studies the time series prediction method based on optimal weighted combination.The specific work is as follows:(1)A time series prediction method of optimal combination model based on particle swarm optimization algorithm is proposed.Firstly,based on the combined prediction method of parameter optimization and component integration,the least squares support vector machine,back-propagation neural network,cyclic neural network and long-term and short-term memory network models commonly used in the field of time series prediction are integrated.Secondly,in the form of constructing objective function,the process of determining the weight ratio of the above four models is equivalent to the optimal solution problem,and the problem is solved based on particle swarm optimization algorithm.(2)Based on the optimal combination model time series prediction method based on particle swarm optimization algorithm and the combination prediction method based on data preprocessing,a combination of clustering and prediction model is proposed to improve the prediction accuracy.Firstly,the subjective and objective weighting theory is used to give weight to each variable of the data.Secondly,the principal component analysis is used to reduce the dimension of high-dimensional time series data,which solves the problems that the traditional clustering algorithm ignores the physical characteristics and unequal importance of each index,and the clustering effect is poor when facing high-dimensional data;Finally,the improved clustering model is combined with the optimal combination prediction model based on particle swarm optimization algorithm,and a time series prediction method based on the optimal weighted combination model is proposed.(3)Taking the measured torque data of permanent magnet synchronous motor and the measured data of wind turbine in Tianjin Shajingzi wind farm as examples,the two algorithms proposed in this paper are verified in turn.Experimental results show that the proposed algorithm effectively improves the accuracy of the original single time series prediction model.