摘要
基于概率统计进行聚类,利用研究所得算法进行特征学习,获得特征权重,可实现特征的选取。此算法有利于特征组合、可屏蔽离散点、不需要预先设定特征的维数和相关性,具有一定人工智能性。经实验验证,此算法适应性较广,有较好应用前景。
Based on the probability of statistics clustering, taking advantages of research algorithm for feature learning, accessing to feature weight can be achieved feature selection. This algorithm is beneficial to the combination of features, it also can be shielded discrete points and do not need to set the characteristics of the dimension and correlation with a certain artificial intelligence. The experimental results show that this algorithm is more adaptable and has a better application prospect.
作者
窦明
吴艳文
Dou Ming Wu Yanwen(Anhui Vocational College of Police officers, Hefei Anhui 230031)
出处
《安徽警官职业学院学报》
2017年第4期126-128,共3页
Journal of Anhui Vocational College of Police Officers
关键词
特征选取
特征评价函数
聚类
feature selection
feature criterion function
cluster