Abstract
The purpose of this research is to develop an optimizing system for a distribution of agricultural products at the wholesale markets using fully-connected neural networks. It is well-known that Boltzmann Machine, which is a specialized type of fully-connected neural networks, can obtain acceptable solutions for NP complete problem, which has no mathematical method for solution. However, this method is available only for selection of problem, selecting an optimal path, and it is not available to determinate analogue data such as quantity of distribution. We, therefore, improved the algorithm of Boltzmann Machine and developed an optimizing system for distribution. This new method enable us to obtain an optimal quantity of distribution which provides high sum of sales and low risk under fixed condition. A change of quantity of agricultural products from a brand product areas, however, influences the prices at the wholesale markets. We further improved the system. After the quantity of distribution was determined, we estimated a change of prices at the wholesale markets. Then, we put the estimated results back to the system and reconsidered the quantities of distribution. Backpropagation neural networks were used for price estimation. This improved system provided more practical solution for distribution.
