一种属性选择方法FS-IV的研究  

Research on a feature selection method FS-IV

作　　者：杨秋洁[1] 胡学钢[1] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《合肥工业大学学报（自然科学版）》2010年第12期1811-1814,共4页Journal of Hefei University of Technology：Natural Science

基　　金：安徽省自然科学基金资助项目(090412044)

摘　　要：数据挖掘所面对的数据常具有属性冗余、包含噪音等特点,使得更注重训练数据质量的分类模型训练周期变长、精度下降。因此,如何选择有效的属性集以约减数据规模,提高分类模型性能具有重要意义。文章将IV模型用于属性选择,提出了基于IV指标的属性选择算法FS-IV,该算法仅需一遍扫描计算出所需的相关统计量,解决了传统属性选择方法处理较大规模数据时空效率不高的问题。实验表明,FS-IV属性选择方法时空性能良好,对冗余、噪音属性均有较好的区分能力,能够有效地约减数据规模。The practical data that mining tasks deal with are often of redundant features and noises,which may lead to lower precision and larger time cost,especially in classification modeling,since high quality data are preferred.Thus,it is of great importance to use those predictive feature data sets for reducing data dimensions and improving the performance of classification modeling.In this paper,a feature selection model FS-IV is proposed based on the information value（IV） index in order to overcome the large time and space cost defect on huge data sets that common selection methods often have.Experimental results show that this model can scan and compute all the coordinated statistics at one time with notable space-time performance and strong ability in identifying redundant features and noises,which effectively reduces the data dimensions.

关 键 词：信息值 属性选择 分类 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]