基于机器学习技术的无线小区负载均衡自优化  

Self-Optimization of Wireless Cell Load Balancing Based on Machine Learning

作　　者：郭华 张东林 徐维华 张航 陈超 GUO Hua;ZHANG Donglin;XU Weihua;ZHANG Hang;CHEN Chao(China Mobile Communications Group Jilin Co.,Ltd.,Changchun 130033,China)

机构地区：[1]中国移动通信集团吉林有限公司,吉林长春130033

出　　处：《移动通信》2022年第4期74-79,共6页Mobile Communications

摘　　要：利用机器学习方法实时对高流量小区进行负载均衡是无线网络智能化的重要课题之一,目前仍无较成熟的应用案例。鉴于此,提出了基于XGBoost以及DNN的无线小区负载均衡参数自优化方法,并分别对均衡切换事件、切换门限进行建模预测。实验结果显示,该方法的切换事件预测准确率达到97.3%,切换门限拟合优度为0.6~0.8,可以高效、精准地实现负载均衡的自优化。Real-time load balancing of high-traffic cell based on machine learning method is one of the important topics of wireless network intelligence,but there is still no mature application case.To this end,a self-optimization method for wireless load balancing parameter is proposed based on XGBoost and DNN.By using this method,the handover events and handover thresholds can be modeled and predicted respectively.Experimental results show that the prediction accuracy of handover events reaches 97.3%,and the fitting goodness of handover threshold is between 0.6 and 0.8,which demonstrates that the proposed method realizes the self-optimization of load balancing efficiently and accurately.

关 键 词：负载均衡 机器学习 网络智能化 无线网络自优化 

分 类 号：TN929.5[电子电信—通信与信息系统]