基于差异指标的概念漂移数据流集成分类仿真

Ensemble Classification Simulation of Concept Drift Data Stream Based on Diversity Measure

集成算法是处理概念漂移数据流的常用方法之一。为了更全面反映基分类器在模型中的整体价值,提出了一种基于差异指标的概念漂移数据流的集成分类算法AE-Div(Ensemble Algorithm for Data Streams with Concept Drift Based on Diversity Measure)。将基分类器的分类准确率和集成差异性进行融合,结合时间因子作为综合度量指标,并根据概念漂移检测情况对基分类器设置不同权重。将AE-Div算法与其它几种使用广泛的概念漂移分类算法在合成数据集与真实数据集上进行仿真。结果表明,AE-Div具有更高的准确率和更好的适应性和稳定性。

Ensemble learning is one of the commonly used methods to deal with concept drift data stream.An ensemble classification algorithm for conceptual drift data streams based on diversity measure is proposed,which is called AE-Div.This algorithm considers both the base classifiersaccuracy and diversity of them,and combines the time factor as a comprehensive evaluation index to measure the value of the base classifiers in the ensemble learning.Then different weights are set for the base classifiers according to the detection of concept drift.Simulation experiments show that the AE-Div has higher accuracy,better adaptability and stability on synthetic data sets and real data sets,compared with other widely used concept drift classification algorithms.