摘要
乳腺微钙化点包含众多属性,由于其中存在的冗余和不相关属性降低了微钙化点病变类型判别的性能。因此,特征子集选择问题成为微钙化点病变类型识别中的重要问题。该文针对传统优化方法用于特征选择的种种缺陷,提出了基于遗传算法的特征子集选择测算法。经乳腺微钙化点特征选择实例分析,证明该方法拥有较强的并行性和寻优能力,在特征选择领域有广阔的应用前景。
Microcalcifications include many redundant and irrelated features, which degrade the microcalcifications classification performance. So, feature subset selection becomes one of the important research issues in the process of microcalcification identification. In view of the deficiencies in traditional combination optimization method, an algorithm of feature subset selection based on genetic algorithm is proposed in this paper. According to the results of practical microcalcification classification example, it is proved that this method possess excellent parallelism and optimization performance.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2007年第1期137-139,153,共4页
Journal of University of Electronic Science and Technology of China
基金
中国博士后科学基金资助项目(2004036063)
关键词
微钙化点
特征子集
遗传算法
特征优化
microcalcification
feature subset
genetic algorithm
feature optimization