A novel approach based on the Gauss-vSVR with a new hybrid evolutionary algorithm and input vector decision method for port throughput forecasting

The prediction of port throughput is very complicate, and its accuracy is affected by many socio-economic factors, particularly affected by their embedded distributed randomness of these factors and mixed noises produced in the processes of data collection, transformation, and calculation. Firstly, in view of the v-support vector regression hybridized with Gauss function (briefed as Gauss-vSVR model), to well solve the nonlinear and mixed noises, this paper uses this model to simulate the nonlinear evolving system of port throughput series. Then, to look for more suitable parameter combination of this model and take into account that GA still suffers from the problems of trapped into local optima and time-consuming, this study integrates the global chaotic perturbation algorithm by using Cat mapping function and local acceleration search algorithm by employing cloud theory, i.e., abbreviated as chaotic cloud genetic algorithm (CCGA), to well determine the parameter values for an Gauss-vSVR model. Additionally, based on the principal component analysis and correlation analysis method, an input vector decision method (namely IVD) is proposed to identify the final input variables for Gauss-vSVR model. Finally, hybridization of IVD and CCGA with Gauss-vSVR model, namely IGvSVR-CCGA, is proposed for port throughput forecasting. Subsequently, the port throughput data and its associate socio-economic factors of two largest Chinese ports, Shanghai Port and Tianjin Port, are employed as practical examples to test forecast performance. The numerical results indicate that the proposed hybrid forecasting model receives more satisfied forecasting performance than other classical prediction models; in the meanwhile, the CCGA algorithm also obtains higher optimal efficiency than other alternative algorithms.