| Due to the attention of economists and computer scientists,the stock market prediction has always been a concern.To establish an effective prediction model,linear and machine learning algorithms have been widely used in the stock market in the past decades.Due to different data sets and different indicators,the performance of a single prediction model is different,therefore,it is not practical to admit that a specific prediction model is the best.In this paper,the VIKOR method(multi-criteria compromise solution ranking method)in the multi-attribute evaluation and decision-making method is applied to combination forecasting.IOWA operator(induced ordered weighted arithmetic averaging operator)is introduced,and a combination forecasting model based on the VIKORIOWA operator is proposed.This paper selects the data of the CSI 300 from the first trading day of January 2018 to the last trading day of December 2020 as the research object and uses 6 indicators such as the opening price,the highest price,the lowest price,and the volume as the input analysis,and the logarithmic closing rate as the output to predict.Five single prediction models-Decision Tree,Random Forest,XGBoost,Support Vector Machine,and Neural Network are studied.Based on the VIKOR method,a set of single prediction models in the integration are calculated to calculate the modified interest ratio,and the calculated modified interest ratio is used as the induction value of the IOWA operator.Taking the minimization of the sum of squares of errors as the criterion,the method for determining the weight of the VIKOR-IOWA operator combination forecasting model is given,to construct the combination forecasting model and carry out the price prediction.The results show that the combined forecasting model is better than each single forecasting model under each forecasting evaluation index,which provides an effective method for forecasting the closing price of the stock index. |
