基于粗糙度的近似概念格动态分类集成学习模型研究与应用

概念格(Galois格)是一种进行数据分类学习的有效工具,然而建格规模庞大使分类效率和准确率受到较大影响.将粗糙度理论应用到概念格分类问题研究中,提出一种新型的近似概念格动态建格和分类挖掘集成学习模型(CACLR).该模型在粗糙度区间根据样本空间分布构建多个相对独立分布且比较精确的近似概念格分类器,能及时消除建格过程中大量与分类知识无关的节点,有效缩减原格规模,融合得到的分类挖掘集成学习模型,具有较好的粗糙分类精度和知识预测学习能力.最后进行CACLR分类集成学习模型在标准UCI数据集中的对比实验,有效验证了该模型的实用价值.

Research and Application of Dynamical Classification Model for Ensemble Learning Based on Approximation Concept Lattice of Roughness

Concept lattice is an effective tool for data classification, but classification efficiency and precision arc effected by its large scale. In this paper, rough sets theory was applied into the classification research of concept lattice, and a dynamical classification model(named CACLR)for ensemble learning based on approximation concept lattice of roughness was put forward. I}his model can constructs some identical approximation concept lattice classifiers of independent distribution and much precision according to the instance spatial configuration at the scope of roughness. And it can eliminate independent nodes in time during approximation concept lattice constructed, reduce the scale of concept lattice effestively. The multi-combination model for ensemble learning has robustness at the accuracy of rough classification and the efficiency of knowledge prediction. In the last part of this paper, the experiments tested on the UCI benchmark data sets were carried on and performance results of were given, which prove the practical value of CACLR model.