Combined prediction model construction method and device

Abstract

The invention provides a combined prediction model construction method and device, and the method comprises the steps: carrying out the K-fold cross verification of a basic model on a training set, and obtaining the prediction value of the training set; after each fold of cross validation, predicting the test set by using the basic model obtained by the fold of cross validation to obtain a predicted value of the test set in the fold of cross validation; taking predicted values of the training set as one-dimensional training vectors; calculating an average value of predicted values of the testset in each fold cross validation to obtain a one-dimensional test vector; training a deep learning model by using the one-dimensional training vector, predicting the one-dimensional test vector by using the deep learning model trained to converge, and calculating an index value of a preset evaluation index of the deep learning model trained to converge; judging whether the index value meets a preset index condition or not; and if yes, forming a combined prediction model by using the basic model subjected to K-fold cross validation and the deep learning model trained to converge. By applying the embodiment of the invention, the prediction precision of the combined prediction model is improved.

Description

technical field [0001] The invention relates to the technical field of combined forecasting, in particular to a method and device for building a combined forecasting model. Background technique [0002] In the actual data forecasting process, due to the interference of many uncertain factors, the single forecasting model often has the problems of insufficient information extraction and low forecasting accuracy, and the combined forecasting model has become one of the research hotspots. At present, most combined forecasting models are linear or non-linear weighted combinations of two models, and the equal weight method or optimal weight determination method is usually used to determine the weights to obtain the combined forecasting model. But in fact, the weight of the combined forecasting model should be flexible rather than fixed, and it has been proved in theory that the combined forecasting method with the optimal weight is not necessarily better than the equal weight or ...

AI Technical Summary

Problems solved by technology

But in fact, the weight of the combined forecasting model should be flexible rather than fixed, and it has been proved in theory that the combined forecasting method with the optimal weight is not necessarily better than the equal weight or single model in terms of forecasting accuracy

Images