分类算法的特征变量选取方法研究.pdfVIP

分类算法的特征变量选取方法研究 摘要 特征选取是模式识别中的关键技术之一。模式识别的最终目的是实现对观察 对象的分类，而分类的基础是模式类的数学表达。在模式识别理论中，模式类是 通过特征来表示的，特征选择的好坏直接影响分类器的性能。在模式识别系统设 计中，特征的确定往往是一个反复的过程，是其中的难点和关键。特征选择往往 有赖于对识别问题的了解，对领域知识有较强的依赖性。特征选择和提取的任务 是如何从许多特征中找出那些最有效的特征，把高维特征空间压缩到低维特征空 间。 本文首先介绍了特征选取的研究背景，发展过程和关键步骤。接着，本文又 粗略介绍了特征选取的一些方法。之后本文详细介绍了T 测试和Fisher 判据进 行特征选取的原理和方法，并设计和实现了相关的算法。最后本文也对两种方法 进行了对比。 关键词：特征选取；T 测试；Fisher 判据； PDF 文件使用 pdfFactory Pro 试用版本创建 ABSTRACT Feature extraction is the elementary problem in the area of pattern recognition．Pattern recognitions aim is to realize the classification of observable obj ect. And classification of is the basis of the mathematical expression mode. In pattern recognition theory, the model is expressed by features. Feature selection will have a direct impact on the performance of classifier. In pattern recognition system design, the characteristics of the sure is often a repeated process, they are the difficulties and key. Feature selection often depends on the knowledge and understanding of the recognition problem, depending on domain knowledge. Feature selection and extraction from the task is to find out how many features that the most effective features, the feature space compression to low high dimensional space dimension characteristic. Firstly, this