摘要
粗集理论通过对原始决策表的约简从而获取规则知识,其核心部分是属性约简。经过约简后的数据更有价值,更能准确地获取知识。文中提出了一种新的启发式属性约简算法,并给出了算法的详细步骤和具体的实验示例。该算法通过不一致计数和互信息增量的计算来衡量属性的重要性,避免了对属性之间随机组合情况的搜索,可以提高求解速度。实验结果表明,相比较于动态约简算法和标准遗传算法,所提出的算法获得的约简属性集更加简洁和高效。
Rough set theory acquires rules knowledge through the reduction of the original decision table,and its core part is reduction of attributes.Data after reduction is more valuable and can obtain knowledge more accurately.Presents a new heuristic algorithm,and proposes the detailed steps of the algorithm.And also an example is given to illustrate the algorithm.The algorithm avoids the search for random composition among attributes via using the inconsistency count and the gain of mutual information criteria to value the significance of an attribute,and increases computing speed.From numerical experiments and comparisons,the algorithm provides more precise and simple reduction of attributes than the dynamic reduction algorithm or the standard genetic algorithm does.
出处
《计算机技术与发展》
2010年第10期16-20,共5页
Computer Technology and Development
基金
国家自然科学基金(70971063)
关键词
粗集
属性约简
启发式
rough sets
attribute reduction
heuristic
