基于频率函数循环重计算的属性约简和挖掘算法研究

针对经典HORAFA启发式约简算法在以属性频率为重要启发信息约简时,往往不能获得最优属性约简集的问题,本文提出了基于属性频率函数循环重计算的改进启发式约简和挖掘算法(BRFA算法)。该算法在已约简属性基础上,进行剩余属性频率函数的循环重计算,直至区分矩阵为空,能大大节省决策表的最小约简时间并能得到所有相对约简。通过实例分析和UCI机器学习数据库实验表明,BRFA算法在属性约简和挖掘方面具有较好的性能。

Research of Algorithm for Attributes Reduction and Mining Based on Re-Calculated Frequency Function

Attributing to the shortage of the typical heuristic algorithm(HORAFA) of the attribute reduction,which can not often get the superior reduction when the attribute frequency is selected as the important heuristic information.The Re-calculation of the Frequency Algorithm(BRFA) is proposed,Which can not re-calculate the frequency function until the discernibility matrix is empty based on the attribute reduction.This algorithm can save some time of acquiring the least reduction in the decision table and get all of the relative reductions.Finally,both the analyzed example and the experimental results by the UCI Machine Learning Date Sets shows that BRFA algorithm is proved more effective in attribute reduction.