Cluster Hedging Mode Based On Moving Hurst Index And Average Grid

In recent years,with the changes in international political and economic situations,the changes in the commodity futures market have shown new characteristics,and the uncertainties affecting their prices have gradually increased,leading to sharp fluctuations in commodity futures prices.For investors,the biggest goal is to maximize revenue and minimize risk.In the future market,cross-species hedging is a typical means of obtaining returns under the premise of effectively avoiding risks.Hedge varieties must have a certain correlation.In previous studies,only two related varieties were used for research,but in reality,certain correlations are not limited to two varieties.Therefore,this paper chooses the cluster hedging related problem as the research object,and uses grid representation,fractal theory and data mining algorithm as research tools to find effective hedging mode as the target.Based on the full study of the achievements of the predecessors,this paper has completed the following work:(1)In the second chapter,several key concepts and ideas in this paper are expounded.The main contents include:the research ideas of cluster hedging,the improvement of grid representation,and the design of variable observer target attributes.Based on this,support for the discovery of the ultimate goal of this article,cluster hedging mode.This paper uses a number of related futures varieties to form a hedging sequence,and explores effective ideas and methods for multi-species hedging through various means.In this paper,the grid representation method is studied.For the needs of cluster hedging research,the moving average grid is supplemented.Compared with the previous grid representation method,combined with the moving average grid,the local features of the hedging sequence can be better expressed.For the time series of the observation point target direction(equivalent to the sample target variable of the general problem),this paper uses a variable polymorphic representation,which can take into account the long-term and short-term reverse changes,which is conducive to effective pattern discovery.(2)In the third chapter,the Euclidean distance is used to select the appropriate futures hedging cluster,and the soybean futures,soybean meal futures and soybean oil futures are finally determined as the three varieties of the cluster hedge.Then,each hedging sequence is calculated,and normality,self-similarity and R/S empirical test are carried out to prove that it has a fractal structure,which can be studied by moving Hurst index.Finally,the V-statistic is used to determine the aperiodic cycle length of the hedge sequence.(3)In the fourth chapter,the method of determining the window length of the moving Hurst index is improved.The general literature uses the aperiodic cycle length as the window of the moving Hurst exponent.In order to better reflect the trend of the hedging sequence,this paper selects other N time lengths as candidate windows in the vicinity of the aperiodic cycle length.The coefficient of variation and the number of predicted reversal points are used to compare the accuracy of the moving Hurst exponent of N+1 different time windows to the trend of the hedging sequence.(4)In the fifth chapter,the selection method of Hurst smoothing period and threshold and the selection strategy of observation point are studied.In order to prevent the influence of individual outliers of the moving Hurst index,the moving Hurst is subjected to 3 to 50 day mean smoothing.In order to avoid the phenomenon that the inflection point is too small when the Hurst index is less than the theoretical threshold value of 0.5,the Hurst threshold range is expanded by 0.1 before and after the study.Calculate the observation point of the moving Hurst index under different smoothing periods and thresholds,extract relevant data,and determine the preferred smoothing period and threshold of each hedging sequence from the angles of the predicted inversion point,the threshold and the hit rate.Finally,according to the preferred smoothing period and the threshold parameter,a selection strategy of the observation point is established,and the feature vector of the observation point is designed.(5)In the sixth chapter,the pattern discovery and verification of cluster hedging is completed.In this paper,In this paper,a hedging sequence and its Hurst index are selected to extract the features of the observation points that conform to the observation point strategy.By using the decision tree data mining algorithm to process the extracted feature vectors,the hedging pattern is generated,and the hedging pattern is manually pruned to retain the pattern that does not violate the market logic and law.Using the same observation point identification and feature extraction strategy,the observation point feature vectors of the other two hedge sequences are obtained,and the found hedging patterns are used for the observation points of the two hedge sequences for verification.In the specific excavation verification scheme in the sixth chapter,the mining model was extracted from the soybean hedge sequence and applied to the soybean meal and soybean oil hedge sequences for verification.It has been verified that the accuracy rate of the pattern in the soybean meal hedge sequence is 83.3%,and the accuracy of the pattern in the soybean oil hedge sequence is 76.9%.This result not only demonstrates the effectiveness of the hedging patterns found in this particular cluster,but also fully demonstrates the effectiveness of the research ideas in this paper.In general,the innovations that support this paper for effective research are as follows:First,the perspective and strategy of cluster hedging are proposed to study the hedging of futures of three or more varieties.Second,the grid representation of the time series and the representation of the point of view are improved.The moving average grid is introduced to better represent the characteristics of the local sequence;the polymorphic strategy is adopted for the future direction of the observation point,which makes the pattern discovery more flexible and effective.Thirdly,the Hurst index calculation and application of various parameter determination strategies have been effectively explored.