The Grey Prediction Model Based On Composition Data And Its Application

Since the introduction of gray system theory in 1981,after more than 30 years of exploration and research,gray system theory has gradually become an important theory and method system for studying and analyzing uncertain systems,and gray prediction is one of the core parts of gray system theory,mainly through The processing of the "partial" known information weakens its randomness,constructs a grey prediction model with partial differential and partial differential,describes the process of system information accumulation,and thus predicts the development trend of the system.With the development of grey forecasting theory,the grey forecasting model is continuously optimized and expanded.At the same time,with the development of data collection technology,many new types of data have also been brought in.Among them,because component data can comprehensively analyze system structure,system component differences and association relationships,and can reveal relative information behind absolute data,There is more and more attention.However,the "summing" feature of component data makes it difficult for traditional prediction models to effectively predict structural data sequences,especially the problem of component data prediction based on poor information features lacks an effective model.Based on this,this paper combines component data analysis and gray prediction model to build a system structure that can predict the development trend of poor information characteristics.The main research contents include:The first is to build a grey prediction model based on component data.Through the existing component data transformation processing method,the "fixed and restricted" component data is eliminated,and the data is converted into Euclidean space,so that the traditional gray prediction model can realize the prediction of the component data,which broadens the application of the gray prediction model range.The second is to improve the prediction effect of the prediction model.Although it is the component data of the poor information system,but because the component data is the overall structural data,it will be affected by the internal components of the system or external uncertain factors,and its change law is not single.Therefore,the first type of model needs to be revised to restore the original Implicit features of component data to improve the prediction effect of the model.In this paper,to solve the problem that the existing gray prediction model cannot directly predict the component data effectively,a GM(1,1)model based on component data transformation is constructed.In the modeling process,component data transformation processing methods are used to eliminate the constraints and constraints of the component data and realize the prediction of the component data sequence.However,considering that the component data is the data that reflects the structure of the system,it will show different changes at different periods due to the interference of the entire system and the various components of the system,and the GM(1,1)model can only effectively predict the change of a single index.sequence.Therefore,this paper considers the random volatility and periodicity of component data,introduces Markov theory and Fourier series,improves the GM(1,1)model based on component data transformation,and proposes based on Markov chain and The modified Fourier series GM(1,1)model makes up for the shortcomings of the GM(1,1)model in predicting single exponential rate data.Finally,the constructed model was used to conduct an empirical analysis of China's energy consumption structure to verify the validity and feasibility of the constructed model.The analysis found that the revised model's prediction effect had improved,and at the same time,the revised GM(1,1)model was used Forecast the energy consumption structure of China in the next 5 years,and put forward corresponding countermeasures and suggestions according to the forecast results.The main points that need to be improved in this article are the following: the existing models can only select the best component data transformation processing method through the application of comparative analysis of actual results,and further research on the mechanism of component data transformation processing is required;second,This article only considers the random volatility and periodicity of the original component data,and the gray prediction model is revised.The revised gray prediction model for other changing characteristics of the component data needs to be further studied and constructed;finally,this paper only selects GM(1,1)To study the model,the application of other gray prediction models to the prediction of component data series also needs further discussion.