Several Kinds Of Interval Combination Prediction Models,Properties And Applications Based On Correlation Coefficients

The combination forecast is the integration of various single forecasting methods,make full use of the advantages of each single forecasting method to the different forecast objects.It can diversify forecast risk across complex system forecasts,reduce the forecasting error,and improve the prediction accuracy,which provide the solid foundation for scientific decision.So the combination forecast has been widely applied in various fields.In the research of combination prediction model,there are two problems worthy of attention and discussion.First,on the premise of the accuracy of each single prediction method,how to allocate the appropriate weight to participate in the combination prediction,that is,how to determine the weights.In order to solve this problem,the current research either adopts the mean weight,or uses the construction error minimum criterion to solve the weight.These methods ignore the interaction between data,resulting in information overlap,and are not widely applicable.Therefore,how to construct an appropriate method to determine the optimal weight has become one of the key contents of the research.Second,complex systems are often faced with uncertain environments,To some extent,the real data cannot fully describe the research object.For example,the temperature in a day,the price of crude oil,etc.,are often expressed by interval numbers.Compared with real numbers,interval data can describe objective information more reasonably,so prediction based on interval information has become another key research content in the field of prediction.To sum up,it is of practical significance to study the determination of the weight of each single forecasting method in combined forecasting based on interval information environment.The following are the research contents of this dissertation:In the first chapter,we briefly introduce the research progress,research methods and main innovations of combinatorial prediction.The second chapter mainly introduces the concept of interval number,the algorithm,the correlation operator and the correlation coefficient of the prediction criteria,which lays a certain foundation for the modeling of the following dissertation.The third chapter,for the interval time series,the COWA operator is introduced to transform the interval data into real data.Secondly,the correlation coefficient in the correlation index is used as the prediction criterion to establish the interval combination prediction model of correlation coefficient based on the COWA operator.The concepts of superiority,inferiority,redundancy,superiority of interval combination prediction are proposed,and the conditions for the existence of non-inferiority combination of the model are analyzed.The cooperative game theory is used to solve the model.Finally,the simulation operation is carried out with the spot price data of West Texas Intermediate crude oil,and the validity of the model is verified from both theoretical and practical aspects.The fourth chapter,compared with the third chapter,carries on different processing to the interval data,the center and radius of the interval number are used to represent the interval value to understand the mean information and fluctuation of the interval.The generalized weighted average(GWA)operator is introduced to build the center and radius correlation coefficient models based on the GWA operator.The multi-objective programming model is transformed into a single objective optimization model by attitude parameters.Some properties of the model are discussed and the conclusion is drawn that the interval combination prediction model is a monotone non-decreasing function of the number of single prediction models in combination prediction.In the fifth chapter,the weights of interval prediction model are improved,the weight coefficients of the single forecasting method in the combined forecasting model are extended from real number to interval number.In this dissertation,a new combination prediction method is proposed,in which both the prediction data and the weight coefficients are interval numbers,and the measure index of interval correlation coefficient is introduced to construct the interval correlation coefficient combination prediction model based on interval weight.The feasibility of practical application of the proposed combination prediction method is illustrated by a concrete example.