Research On Bid-Sections Partition In Government Drug Procurement

As a main purchasing model of medical organizations, centralized bidding and purchasing system of durg has achieved a significant effect in lowering drug prices and improving procurement efficiency. Under the promotion of the relevant departments, centralized bidding and purchasing will become the only national drug procurement mode. However, this mode is still in an exploratory stage, the theory obviously can not keep up with the practice and application. One of the biggest problems is that the organizers take bid subject strictly according to the specific property of drugs, in other words, they do not have a reasonable and scientific approach to partition drugs during the bidding process, which results in high time and management costs for related players and unreasonable bid prices. Thus, this paper aims to investigate this problem and design an efficient method for partitioning bid-sections in the bidding procurement of drugs.Firstly, this paper analyzes the main mode of bidding and purchasing system of durg from the application perspective, on the basis of which, introduces the combinatorial auctions theory and economics of Scale theory to demonstrate the feasibility and potential advantages of partitioning bid-sections in drug procuement. Secondly, based on the theory study, this paper analyses the purpose of partitioning bid-sections and decides to use the K-prototypes clustering algorithm in designing the partitioning method. The design Process includes the following steps:determining the target of partition, screening dissimilarity measure, improving the Clustering Algorithm, designing the operation process and programming. In order to improve the Clustering algorithm, this paper use the Density-based method in the selection of initial prototypes and the iteration of clustering center. What's more, We Set the parameters from three dimensions: the Weight between value attribute and classification attribute, the weight between classification attributes and the distance of classification attribute. We determine the Weight between classification attributes referencing the W-k-prototype algorithm, and use the level set method to compete the set of the distance of classification attributes. Finished the design of partitioning method, this paper programs the improved algorithm and its evaluation method with C # .To examine the validity of the method, this paper makes a simulation experiment with the real case of drug bidding procurement in Qujing city Yunnan province, 2009. Each bid-section in the partitioning result has a very high purity in the formulations and classifications of drugs. The size of bid-sections is similar and reasonable, the proportion of clusters whose average lowest price index that is less than 0.3 reaches 90%, which indicates that the bidding advantages of tender companies has been fully embodied.The result shows that the partitoning method is feasible and the application of partitioning bid-sections has an important impact on the improvement of the centralized bidding and purchasing system of durg.