基于扩张矩阵的渐进式特征子集选择算法

特征子集选择问题一直是人工智能领域研究的重要内容,特别是近几年来,特征子集选择的算法研究已经成为机器学习和数据挖掘等领域的一个研究热点。该文在扩张矩阵的基础上提出了类扩张矩阵的概念,并将加权的期望信息和不一致错误率函数应用于特征子集的选择,实现了具有噪音处理功能的渐进式特征子集选择算法———IFSS_EM,实际领域的实验结果表明:IFSS_EM算法具有运行效率高、选择特征较具有代表性的优点,从而使其能够较好地应用于实际领域。

Incremental Feature Subset Selection Algorithm Based on Extension Matrices

Feature Subset Selection(FSS)problem has long been active research topic in the area of Artificial Intelligence.In recent years,FSS has become focus in the fields of Machine Learning,Pattern Recognition and so on.In the paper,a new concept ,Class Extension Matrix is proposed,which is derived from Extension Matrix.The weighed expected information and inconsistency error rate are applied to the selection of feature subset.A new feature subset selection algorithm tolerating noise and incremental,Incremental Feature Subset Selection Algorithm based on Class Extension Matrices(IFSS_EM)is designed and implemented.Empirical results in the real-world datasets show that high efficiency can be achieved and more representative features can be also obtained for IFSS_EM.This implies that IFSS_EM can be applied to the real-world datasets effectively.