OWA-weighted based clustering method for classification problem
OWA-weighted based clustering method for classification problem
复制标题
DOI：
10.1016/j.eswa.2008.06.013
复制
发表时间：
2009-04
期刊：
影响因子：
--
通讯作者：
Ching-Hsue Cheng;Jia-Wen Wang;Ming-Chang Wu
中科院分区：
文献类型：
--
作者：
Ching-Hsue Cheng;Jia-Wen Wang;Ming-Chang Wu
信息分类在决策问题中发挥着重要作用。随着信息技术的进步，大量的信息存储在数据库中。许多任务都是在分类问题中以高复杂性和高维度来解决的。因此，本文应用有序加权平均（OWA）算子将多属性数据融合为单属性的聚合值，并对聚合值进行聚类以用于分类任务。该方法包括四个步骤：（1）使用逐步回归来选择和排序重要属性，（2）利用OWA算子从多属性数据中获取单个属性的聚合值，（3）通过K-means方法对聚合值进行聚类，（4）预测测试数据的聚类。在验证和比较中，通过所提出的方法进行了三个数据集：（1）鸢尾花、（2）威斯康星州乳腺癌和（3）关键绩效指标数据集。解决了复杂度高、维数高的问题，分类准确率高于现有的一些方法。
Information classification is an important role in decision-making problems. As information technology advances, large amounts of information stored in database. Many tasks are worked out in high complexity and dimensionality in classification problem. Therefore, the paper applies ordered weighted averaging (OWA) operator to fusion multi-attribute data into the aggregated values of single attribute, and cluster the aggregated values for classification tasks. The proposed method consists of four steps: (1) use stepwise regression to select and order the important attribute, (2) utilize OWA operator to get aggregated values of single attribute from multi-attribute data, (3) cluster the aggregated values by K-means method, (4) predict the clusters of testing data. In verification and comparison, three datasets: (1) Iris, (2) Wisconsin-breast-cancer, and (3) Key Performance Indicators datasets are conducted by the proposed method. The problems of high complexity and dimensionality are solved and the classification accuracy rate is higher than some existing methods.