基于种群协同进化的分类判定树构造算法

提出一种基于两种群协同进化的分类判定树构造算法，该方法充分利用协同进化的强搜索能力和渐进学习等特点，通过设计两个协同进化的种群：一个表示选择的属性子集，另一个表示如何构造判定树，保证在搜索曩优判定树的过程中同时对属性子集选择、判定树的构造进行综合优化，最终获得一个较好的分类判定树。作为实验验证，我们把新算法应用到一个困难的真实问题一胸癌诊断，结果表明了新算法的有效性。和其它算法的实验结果比较，新的分类方法比C4.5和文1]方法构造出更好的判定树，即去除了多余的属性、具有更高的分类精度。

Using Coevolution for Building Decision Tree

This paper puts forward a construction algorithm of decision tree based on two colony coevolution. This algorithm utilizes fully the strong searching power and stepwise learning ability of coevolutionary computation and guarantees to optimize the two problems synthetically in the search of optimal decision tree by designing two simultaneously evolving populations: one represents subset of features selected, the other represents how to build a decision tree, and then obtains a better decision tree for classification. In order to prove the validity of this algorithm, we apply it to a hard and real problem the diagnosis of breast cancer, the result shows the effectiveness of this new algorithm. Comparing with other methods, this new method constructs a better decision tree than that produced by C4.5 and that produced by the method of . That is to say our new algorithm removes redundant features and gains better precision of classification.