面向装备作战仿真数据流的改进型贝叶斯分类方法研究  

Study of improved Bayesian classification method for data stream of equipment combat simulation

作　　者：曹波伟[1] 薛青[1] 汤再江[1] 

机构地区：[1]装甲兵工程学院仿真中心,北京100072

出　　处：《计算机工程与科学》2013年第12期167-172,共6页Computer Engineering & Science

摘　　要：针对装备作战仿真数据流的无限流入和概念漂移现象影响分类模型准确度的问题,提出了一种基于权值窗口的增量学习型朴素贝叶斯分类算法。该算法通过在装备作战仿真数据流上建立权值窗口,以充分利用历史时间数据的后验信息学习新分类模型,目标是降低装备作战仿真数据流的无限流入和概念漂移现象对其分类模型准确度的影响,提高朴实贝叶斯分类模型的准确度。数值实验说明了该算法的有效性,并且其在分类性能、分类准确率、分类速度上优于同类算法。According to the problem of the endless flowing of data stream of Combat simulation and the affections of concept phenomenon on its classification model accuracy,an incremental learning simple Bayesian classification algorithm based on Weight window is proposed.The algorithm build a weight window on the data stream of equipment combat simulation in order to learn a new classification model with taking full advantage of the historical time data.The goal is to reduce the unlimited inflows of data stream of equipment combat simulation and mitigate the impact of the concept drift on the accuracy of the classification model,thus improving the accuracy of the simple Bayesian classifier model.Numerical experiments also prove that the algorithm is effective and outperforms other similar algorithm counterparts in terms of classification performance,classification accuracy rate and classification.

关 键 词：装备作战仿真 数据流分类 概念漂移 贝叶斯 

分 类 号：TP391.99[自动化与计算机技术—计算机应用技术]