A novel composition forecasting model based on choquet integral with respect to Q-measure

In our previous works, all of the multivalent fuzzy measures which based on the p-measure and additive measure always do not contain the well-known fuzzy measure, lambda-measure. In this paper, based on p-measure and lambda-measure, an improved multivalent fuzzy measure which contains lambda-measure, called Q-measure, is proposed. Based on a new fuzzy density, O- density, and the Choquet integral respect to Q-measure, a novel composition forecasting model composed of the time series model, the exponential smoothing model and the GM (1,1) forecasting model is proposed as well. An experiment with real data by using the 5 fold cross validation mean square error is conducted. The performances of the Choquet integral composition forecasting model with the Q-measure, LE-measure, the L-measure, the Lambda-measure and the P-measure, respectively, are compared with the ones of the ridge regression composition forecasting model, the multiple linear regression composition forecasting model and the traditional linear weighted composition forecasting model. The experimental results show that the Choquet integral composition forecasting model with the proposed Q-measure and the O-density has the best performance.