摘要
XGBOOST是一种集成学习算法,通常采用网格法调参,参数优化幅度有限;提出连续量子粒子群算法,并将其引入XGBOOST的调参过程,得到全局优化参数,从而提升XGBOOST算法性能;同时将该算法应用于电信精准营销案例——运营商用户换机预测场景中,得到了基于连续量子粒子群算法调参的XGBOOST用户换机预测模型,与基于网格法调参的XGBOOST相比,该模型获得了更高的用户换机的预测准确度。
XGBOOST is an ensemble learning algorithm,which usually uses grid method to adjust parameters and the parameter optimi- zation scope is limited. The algorithm of parameters adjusting based on Continuous Quantum-inspired Particle Swarm Optimi- zation is proposed,which improves the performance of XGBOOST. This model has been applied in a case of telecom precision marketing,the prediction of user equipment replacing.Compared with XGBOOST based on Grid parameter adjustment method, the model achieves higher prediction accuracy of user equipment replacing.
作者
成晨
程新洲
张恒
韩玉辉
Cheng Chen;Cheng Xinzhou;Zhang Heng;Han Yuhui(China Unicom Network Technology Research Institute,Beijing 100048,China)
出处
《邮电设计技术》
2018年第10期20-24,共5页
Designing Techniques of Posts and Telecommunications
关键词
连续量子粒子群
XGBOOST
机器学习
电信精准营销
Continuous quantum-inspired particle swarm optimization （CQPSO）
XGBOOST
Machine learning
Telecom precision marketing
