基于AdaBoost的集成分类器在电信增值业务中的应用

Ensemble Classification Based on AdaBoost Algorithm and Its Application to Value-Added Services of Telecommunication Industry

为了解决数据挖掘技术较难有效地在电信行业挖掘出潜在增值业务用户的问题,针对当前单分类器分类精度低这一不足,提出一个基于BP神经网络与AdaBoost结合的集成分类器模型。选用BP神经网络作为基分类器,通过Ada-Boost算法进行T轮迭代,每次迭代增加错分样本的权重,最终通过投票产生强分类器。通过对中国电信某地市用户消费数据进行实例仿真,证明该模型能有效地提升分类精确度,分类精度达到76.7%,并且拥有不错的鲁棒性,为以后的研究工作提供了新的研究思路。

In order to solve the problem that the potential users of value-added services can＇ t be extracted effectively by DM technology, a new ensemble prediction model based on BPNN and AdaBoost algorithm was proposed to overcome single classification＇ s low precision. In the model, BPNN was adopted as base classifier, whose precision was improved by AdaBoost algorithm. With the experiment on the users＇ data of ticket, it showed that the model, which had a good robustness, improved the accuracy of classification, which provided a new research approach for future research.