Abstract: Aimed at the contradiction that the coal quality information is scattered and the quality-valuation policy requires the precision of its quality, which is existing in coal sales currently, a new K-Means clustering algorithm optimizing the selection of the initial clustering center is proposed, and a detailed analysis and data mining on the data of the coal quality inspection produced by a large-scale coal enterprise and its alliance customers is conducted, in order to gain the statistical rules of the traders’ quality inspection behaviors, and grade credibility hierarchies of the traders on both sides, thus transforming the uncertain quality indicators into the deterministic evaluation of the traders’ quality inspection behaviors. On one hand, it can provide reference for the coal enterprises to divide the responsibility of quality inspection reasonably when the quality disputes of the traders occur, and supervise and guide the quality inspection managements of mines and customers at the same time; on the other hand, it can provide ancillary support for the sales decision of the coal enterprises.

Key words: credibility, supply chain, clustering center, quality inspection, K-Means clustering

摘要： 针对目前煤炭销售中存在的煤炭质量信息分散性与质量计价政策对其质量精确性要求的矛盾，提出了采用对初始聚类中心优化选取的K-Means聚类算法，对大型煤炭企业及其联盟客户煤炭质量检验数据进行系统分析和挖掘，获得双方质量检验行为统计规律，进行交易双方信誉度等级划分，将不确定性的质量指标转化为确定性的交易者质量检验行为评价。该研究一方面可为煤炭企业在发生交易者质量纠纷情况下合理划分质量检验责任提供参考，监督和指导矿井和客户的质检管理工作;另一方面可以为煤炭企业提供销售决策辅助支持。

关键词: 信誉度, 供应链, 聚类中心, 质量检验, K-Means聚类