AN INTEGRATED RS-SVM MODEL USING GENETIC ALGORITHM AND CROSS VALIDATION FOR PREDICTING THE POTENTIAL DEMAND OF AGRICULTURAL SUPPLIES

摘要

In the distribution process of agricultural supplies, there are some factors impacting the potential demand, so precise supply orders are often not known in advance. Meanwhile, the dimensions of these impact factors are different and measured with various criteria. Thus, it is difficult to determine the indexes to predict the potential demand of agricultural supplies. In order to cope with the problem, this paper identifies and presents 16 specific indexes of the potential demand prediction depending on four aspects, i.e., customer, enterprise, society and logistics. Then, a hybrid algorithm, that is, RS-SVM, is proposed based on rough set (RS) and support vector machine (SVM). In the proposed algorithm, genetic algorithm (GA) as one of attribute reduction approaches in RS is chosen and used to achieve the attribution reduction, by which the number of indexes is reduced from 16 to 6. Then, SVM is designed and performed to predict the potential demand of agricultural supplies in the agricultural distribution. Finally, a numerical example is given and the comparison results demonstrate the effectiveness and applicability of the proposed algorithm.