作 者: ;
机构地区: 广东司法警官职业学院
出 处: 《科技通报》 2013年第8期79-81,共3页
摘 要: 特征选择是数据挖掘领域中有效的数据预处理算法,该算法能够提取对训练模型有价值的特征属性。传统的特征选择算法往往是针对小规模数据。针对大规模数据,传统的特征选择算法不能有效地运行,计算效率很低。本文针对海量高维数据,提出了基于Hadoop平台的分布式特征选择算法。该算法能够有效地完成特征属性的提取工作,并且,算法的执行效率很高。实验结果显示,本文提出的算法具有很高的加速比。 Feature selection is an effective data preprocessing algorithm in data mining area, and this algorithm could ex- tract attributes which are valuable for training model. Traditional feature selection algorithms usually focus on small scale data, however, focusing on big scale data, traditional feature selection algorithms could not run effectively, and the effi- ciency of them are very low. In this paper, focusing on big scale, high dimensional data, we propose a distributed feature selection algorithm based on Hadoop platform. This algorithm could complete the feature extracting work effectively, and it has good efficiency. The experimental results show that the algorithm in this paper has good speed-up.
领 域: [自动化与计算机技术] [自动化与计算机技术]