Port Throughput Forecasting by Using PPPR with Chaotic Efficient Genetic Algorithms and CMA

The forecasting accuracy of port throughput is influenced by many complicate socio-economic factors, especially by nonlinear and fluctuation signal included in the history sequence of port throughput. To deal with the nonlinear and fluctuation signal in the history sequence, this paper introduces the parameter projection pursuit regression (PPPR) method to the port throughput prediction field. In the process of modeling PPPR, the orthogonal Hermite polynomial is used to fit ridge functions therein, and the least square method is employed to determine the polynomial weight coefficient c. Then, in order to optimize projection direction a of PPPR model efficiently, combining the cat mapping function, accelerating evolutionary mechanism and the GA algorithm, the chaotic efficient genetic algorithms (CEGA) is proposed for optimizing projection direction a. Finally, a port throughput forecasting approach, adopting CEGA algorithm to optimize the optimal projection direction a in inner layer while optimizing the ridge function number M in outer layer, employing correlation analysis method (CAM) method to determine the input variables, namely PPPRCEGA, is proposed. Subsequently, this study compiles numerical experimentation to evaluate the feasibility and performance of the proposed approach. The empirical results indicate that the proposed forecasting approach improve the precision of the forecasting results of port throughput.