一种用于特征选择的禁忌搜索算法

A TABOO SEARCH ALGORITHM FOR FEATURE SELECTION

特征选择问题是机器学习和模式识别中的一个重要问题,特征的优劣直接影响分类器的设计和性能。首先介绍了禁忌搜索的基本原理,然后将禁忌搜索方法用于特征选择,并给出了新算法的实现方法及步骤。仿真实验表明,该方法在求解的效率和解的质量方面都达到了令人满意的效果。

Feature selection is an important issue in the fields of machine learning and pattern recognition.The effectiveness of feature directly affects the design and performance of the classifier.In this paper we firstly introduce the general idea of tabu search,then apply the tabu search to feature selection,and provide the implementing approach and procedure of the new algorithm.It is shown by the simulative experiment that this method achieves satisfying effect in both solving efficiency and solution quality.