基于信息熵的连续属性自动聚类算法

Automatic Clustering Algorithm for Continuous Attributes Based on Entropy of Information

基于信息熵的有关理论 ,提出了一种新的连续属性的自动聚类算法。首先介绍了 Shannon熵的概念及其两个重要的定理 ,基于信息的不确定测度 ,提出了一种 Shannon熵的准则函数 φ,并且指出了该准则函数必须满足的 6条原则。其次 ,基于该准则函数 ,引出了一种针对单个连续属性自动聚类的 FUSINTER算法。由于实际信息系统中有多个连续属性 ,这就需要对多个连续属性分别使用 FUSINTER算法进行离散 ,并且要求最终保证整个信息系统离散后是相容的和一致的 ,而且各个属性拥有较少的分割区间。最后 ,本文以干线飞机外形参数的变化趋势与其更新换代的关系来说明文中提出的连续属性离散化过程 ,并展示了该聚类算法的有效性。本文提出的方法可以用于机器学习或数据挖掘的数据前处理。

Based on theory of the entropy of information, a ne w method of automatic clustering algorithm is presented. Firstly, this paper int roduces the concept of Shannon entropy and its two important theorems. According to uncertainty measures of information, a criterion φ based on Shannon′s e ntropy is presented, and the criterion φ must be satisfied to six axioms. Then based on the criterion φ, FUSINTER algorithm aimed to the single con tinuous attribute, is introduced. In fact, an information system alway s has many continuous attributes, so FUSINTER algorithm will be used several times to scattering the many continuous attributes. At the same time, after aut omatic clustering, information system is compatible and coincident, and each of the attributes has fewer intervals. An example is given to prove the efficiency of the clustering algorithm.