Monitoring Data Processing Method And Application Based On Combination Prediction Model

Data processing,analysis and prediction are important contents of deformation monitoring,and the estimation and prediction of the development trend or deformation value of deformable body is one of the important application values of deformation monitoring.The research method of combination forecasting has been put forward since the sixties of last century.After decades of development,it has become one of the important research directions of forecasting discipline,and it has been widely used in various industries,such as commodity price forecasting,economic forecasting,power load forecasting and so on.There are also related applications in deformation monitoring data processing,such as deep foundation pit displacement deformation prediction,geological hazard early warning,foundation settlement monitoring,dam displacement monitoring and so on.By combing the important literature at home and abroad,it is found that in the field of monitoring data processing,there are more researches on the establishment of traditional fixed weight coefficient combination forecasting model based on some optimal criteria,but there are few studies on other combination theories and methods.Therefore,this paper mainly discusses the combination forecasting model from two aspects,and studies the combination forecasting method and prediction accuracy in monitoring data processing.On the one hand,based on the traditional combination forecasting method,some calculation methods or prediction algorithms in the model are improved to improve the prediction accuracy of the displacement change of the deformed body;on the other hand,it is a preliminary exploration of the dynamic(variable)weight coefficient combination forecasting method,which is difficult in the research of combination forecasting discipline.The specific work is as follows.First,the model and prediction method are improved.(1)For the optimal weight coefficient combination model when combining different single prediction models,the problem of excessive partial weight occurs due to the different fitting accuracy of the model.this paper proposes A combination forecasting model of weight coefficient based on local optimum,namely,take the sum of squares of residuals of each prediction model from time t₀ to time N as the objective function,then determine the local optimal weight of the model by using the least square criterion,and use it as the weight of the combinatorial forecasting model.The effectiveness of this method is verified by the measured water level monitoring data,compared and analyzed with the combination forecasting model of optimal weight coefficient and the measured data,the result shows that this method can solve the problem of excessive bias weight and improve the prediction accuracy of the combination mode.(2)In view of the current problem of the accuracy of landslide displacement prediction is not high for the single model,the paper starts from the perspective of correlation combination prediction,combined with induced ordered weighted averaging operator and proposed a combination model of landslide displacement correlation based on induced ordered weighted averaging operator.According to the grey metabolism idea,a metabolism prediction method based on the combined model of landslide displacement correlation based on the induced ordered weighted averaging operator is proposed.By introducing the fitting prediction evaluation index to evaluate the fitting and prediction effect of the model.Through the results of the example verification show that the fitting evaluation indexes of the landslide displacement correlation combination model based on the induced ordered weighted averaging operator are better than the selected single item model and the correlation combination model,and the prediction index is also improved compared with the original method.Second,research on dynamic(variable)weight coefficient combination forecasting model.(1)Research on linear combination forecasting model of moving weight coefficient of landslide position based on least square criterion.In view of the fixed weight coefficient of combination forecast model in multiphase forecasting problems in precision drops too quickly,this paper based on the least squares criterion of linear combination forecasting model,combining with the metabolism of ideas,put forward a kind of based on the least squares criterion of landslide displacement dynamic weight coefficient linear combination forecast model,with the introduction of prediction evaluation index to evaluate the prediction effect of this model.The results show that the linear combination prediction model based on the least square criterion is superior to the single prediction model and the ordinary linear combination prediction model.(2)Study on Metabolic combination forecasting model of landslide displacement based on correlation coefficient.For prediction accuracy of combination forecasting model drops too quickly in the landslide displacement,this paper starts from a combination forecasting model based on correlation coefficients,and proposed a metabolic combination forecasting model of landslide displacement based on correlation coefficient combined with the idea of metabolism,and defined and derived the corresponding calculation formula.By introducing the prediction evaluation index system to evaluate the prediction effect of the model.The result of the example of landslide displacement shows that the prediction evaluation index of the metabolic combination forecasting model of landslide displacement based on correlation coefficient is better than that of the single prediction model and the combination forecasting model based on correlation coefficient.(3)The metabolic combination prediction method of landslide displacement based on entropy weight hierarchy structure.In this paper,the analytic hierarchy process and metabolism theory are integrated into the combination prediction model,and proposed a method of metabolic combination prediction of landslide displacement based on entropy weight hierarchy structure.First,the combination prediction model is constructed from the two indicators of"fit accuracy"and"development correlation",and calculated the combination weight of each single prediction model by the combination of least square model and the grey relational analysis.Afterwards,according to the combination weight of single prediction model,a hierarchical structure is constructed by index through the use of entropy method,and the final weight of each single prediction model is determined by calculating the entropy weight of each index.Then,metabolism theory is introduced to eliminate stale data and add new data by way of self-metabolism to transform weight from fixed weight coefficient to dynamic.Finally,the method proposed in this paper is verified by taking the displacement monitoring of a dam as an example.The experimental results show that the method proposed in this paper has certain stability and effectiveness.