容忍噪音的扩张矩阵启发式算法研究

A Noise-Tolerant Heuristic Algorithm for Extension Matrix

现有扩张矩阵算法多为建立在理想数据基础上的,而实际的应用领域中不可避免地存在噪音数据,这样致使其在实际的应用中很难得到令人满意的结果。文章对原有扩张矩阵理论进行扩充,提出扩张矩阵集的概念,并在此基础上给出了一个容忍噪音的扩张矩阵启发式算法(NCV)。实际领域的实验结果表明:NCV算法能够得到较为简单而精确的规则,并且较好地解决了实际领域中存在的噪音问题。

Extension Matrix is constructed from noise-free datasets.However it is inevitable noises exist in the real-world applications,which make it not be able to obtain better results for the algorithms based on Extension Matrix.This paper proposes a Generalized Extension Matrix,which is the extension of Extension Matrix.A new heuristic algorithm based on Generalized Extension Matrix,NCV is also given.The empirical results show that NCV can obtain simpler and more precise rules and handle noises in the real-world datasets effectively.