| Metal futures price forecasting has important applications for investors,traders,government departments and other market participants.Traditional forecasting methods based on statistical and economic theories have certain limitations,while support vector regression models(SVR)perform better in handling nonlinear and high-dimensional data with the advantages of simple modeling and less time consumption,but the penalty coefficients,insensitivity coefficients and kernel parameters of SVR can significantly affect the forecasting performance.In the article,an improved honey badger algorithm(IHBA)is proposed and used to optimize the penalty coefficients,insensitivity coefficients and kernel parameters of SVR,and the IHBA-SVR algorithm is proposed.Meanwhile,in order to improve the comprehensiveness of modeling information in metal futures price forecasting,various methods are used to construct shallow and deep features and combine them with the IHBA-SVR algorithm to build an integrated forecasting model for metal futures prices.The specific work is as follows:(1)Aiming at the shortcomings of honey badger optimization algorithm,such as low population diversity,easy to fall into local optimum,and premature convergence,an improved honey badger optimization algorithm based on Cubic chaotic map,sinecosine algorithm and t-distribution variation disturbance is proposed.The superiority of the improved honey badger optimization algorithm is verified on 11 benchmark functions,indicating that these improvements can greatly improve the optimization performance of the original honey badger optimization algorithm.(2)First,the construction method of multi-class features for metal futures prices is proposed: the sliding window method is used to construct shallow features,and the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and the bidirectional gated recurrent unit(Bi-GRU)are used to obtain two deep features,respectively.Then,the IHBA-SVR is used to model the fused shallow and deep features,and an integrated model for metal futures price forecasting,CEEMDAN-Bi-GRUIHBA-SVR,is proposed.Finally,the closing prices of three metal futures products are selected,and CEEMDAN-Bi-GRU-IHBA-SVR is compared with other forecasting models for experiments,and the results show that the evaluation indexes of CEEMDAN-Bi-GRU-IHBA-SVR are all better than other forecasting models,which verifies the effectiveness of combining multiple types of features and the strong superiority-seeking ability of IHBA.CEEMDAN-Bi-GRU-IHBA-SVR has outstanding forecasting ability and can provide reference for related enterprises,financial institutions and investors. |
