摘要
阐述了传统遗传算法的基本思想、原理和步骤及其在数据挖掘(规则集发现)中的应用,给出了基于遗传算法的知识规则挖掘算法的基本思想和关键问题,包括知识规则表示、适应度函数定义等,继而提出多种群并行进化结构,利用精英重组策略,产生池进化模型以及自适应参数的手段调整并行遗传算法进行数据挖掘。在算法具体实现过程中,采用了动态变异交叉概率等方法,有效避免了并行遗传算法中早熟现象的发生。以北美香菇数据为例,进行并行遗传算法挖掘分类规则,实验说明了该算法在发现和进化规则方面的有效性。
Presented the traditional genetic algorithm, the principles and the processing steps of the data mining （rule set discovery）. Then proposed the basic thinking and the key problem of this algorithm, including the representation of the rule and the definition of the fimess - function etc. Then used parallel genetic algorithm which added the application of multiple and parallel evolutionary group structure, the elite reorganization strategy, the productive-pool strategy and adaptive parameter adjustment methods for data mining. Meanwhile, in the process of algorithm realization, used the dynamic variation of crossover probability to effectively prevent the genetic algorithm phenomenon of precocious puberty. Finally, used the North - American- mushrooms data as examples to try to use genetic algorithms finding classification rules, and successfully proved the validity of the algorithm in discovering and evoluting the rules.
出处
《计算机技术与发展》
2008年第8期137-139,181,共4页
Computer Technology and Development
基金
国家自然科学基金(60273043)
安徽省教育厅自然科学基金重点科研项目(2006KJ013A)
