Logistics Forecasting Methods And Application Study Based On Optimization Algorithm

As an important branch of the modern heuristic optimization techniques, Genetic Programming (GP) is a global optimizing method generated from traditional genetic algorithms and a main part of well-known paradigms in evolutionary computation. Using the Least Square Method (LSM), this paper extends the ordinary GP, and obtains a novel nonlinear forecasting model, noted as Genetic Programming based on Least Square Method (GP-LSM). According to the basic idea from integration technology, this paper also presents a nonlinear integrated forecasting model based on our GP-LSM. Compared with ordinary GP model, our two new models proposed in this paper possess four merits:a new form of individual expression, a new measure of the fitness function, the adaptive replication strategy, and the dynamic crossover and mutation strategy.As one of reasonable applications of our algorithms developed in this paper, we empirically predict the container throughput of Tianjin Port and Qingdao Port using the monthly data. A comparison between our methodology and several classic modelling techniques is presented in the section of empirical study. Broadly speaking, GP-LSM algorithm and its corresponding integrated prediction model, provide the clear and precise function expressions which can explicitly instruct our proper application of the estimated models, and improve prediction accuracy and direction prediction accuracy to some extent. In the end, we demonstrate the respective predictions of the container throughput of both ports in the first six months of2014in the light of different forecasting models and TEI@I methodology.