Research On Bid-Section Division Of Foreign Periodicals Based On SOM Neural Network

Bidding is introduced to the procurement of foreign periodicals of university's library. Because the price of periodicals increases year by year, meanwhile the sort of foreign periodicals involved in the purchase is various, the purchasing is complicated. It is an important problem which is needed to solve for university's library that how to conduct reasonable bid-sections for the foreign periodicals by scientific method, improve the efficiency of bid-sections and reduce the procurement cost, advance the quality of service, so as to optimize the procurement results.Self-organizing neural network theory,which was used to cluster and the related literatures were firstly analized in this paper, and it's determined to use the method of SOM neural network clustering to achieve the division of bid-section.Then It's gave design proposal of bid-section division,which is based on som neural network,and at the same time,it's gave the algorithm of two indexes for evaluation of the effectiveness of clustering.Namely indexes of SD index and average lowest price index.Then it is the application of bid-section division combined with the data of foreign periodicals purchasing of Harbin Institute of Technology library in 2010.Here the process of the SOM neural network clustering is introduced in detail,and the clustering results is analyzed.It showed the selection process of the best of clustering results to make up shortage of determining the clustering results mainly on the experience.It's found that SOM neural network is practical significance and feasibility in bid-section division of the purchasing of foreign periodicals of library.Finally, it's compared the optimal program of SOM neural network clustering to the actual program of bid-section division of foreign periodicals purchasing in 2010,and It's found that SOM clustering method can be obtained bid-section division automatically, quickly and scientificly ,the improvement of existing method ,which was reliedmainly on the manual and experience . Compared with the program of k-means clustering,It's found that SOM clustering method was more suitable for solving the problem of bid-section division of foreign periodicals purchasing.After the comparison of scheme obtained by improved K-prototypes which is proposed by previous high of the line of Duty, we found that SOM neural network clustering is another solution in bid-sections for library's procurement of foreign periodicals, moreover the method has some advantages.