Design Algorithm for Forecasting Model of Transport-Logistic Activity on the Basis of Neuro-Fuzzy Networks

Abstract

Purpose: To consider the current situation in container transportation market of Leningrad Region. To show the need in the design of new forecasting models. To justify the use of neural fuzzy networks. To consider existing limitations of their application. To suggest the ways to build predictive neural fuzzy models for evaluating promising quantitative indicators of logistics activities. Methods: ANFIS, R-ANFIS, Fuzzy-Partitions, SCRG, GD, LSE, PSO, ABC, FA. Results: The necessity of adaptation of neural fuzzy networks for making forecasts in the field of logistics is pointed on, the structure of a promising model is proposed which takes into account the use and analysis of variety of methods on structural and parametric identification. The shortcomings of particular methods of structural and parametric identification, which affect the accuracy of being obtained forecasts, are indicated. Practical importance: The importance of design of accurate predictive models for key quantitative indicators of logistics activities is shown. The factors influencing the implementation of transport and logistics activities are given which of, reliable forecast is impossible in the situation of limited time and resources. For consideration, the article proposes the existing methods of parametric and structural identification of fuzzy neural networks. An algorithm for adapting existing methods to the use in a transport and logistics process is proposed.