逻辑回归算法在通信GPRS业务中的应用

发布时间：2017-12-26 19:05

本文关键词：逻辑回归算法在通信GPRS业务中的应用　出处：《西安工业大学学报》2016年11期 　论文类型：期刊论文

【摘要】：为了对通信GPRS业务用户的流失情况进行有效的预测,对预测中常用三种算法的优劣势进行了比较,选取能很好处理0/1分类问题的逻辑回归算法,并基于逻辑回归算法建立了GPRS业务流失预测模型.提取了广东省移动公司GPRS业务用户流失概率最高的前5%和10%用户,通过查准率、查全率和提升率这三个指标对该模型进行检验,发现该模型定位流失用户的准确率和确定流失用户的覆盖率都是相当高的,说明其能对GPRS套餐使用客户的流失情况进行有效地预测.另外,通过把利用逻辑回归算法与利用决策树算法建立的预测模型的应用效果进行了对比,结果充分说明了利用逻辑回归算法建立的GPRS业务流失预测模型在实际应用中更具优越性.最后,根据该模型解在决实际预测问题中的效果,进一步验证了其具有很强的实用性.
[Abstract]:In order to loss of communication of GPRS service users effectively predict, forecast of three kinds of commonly used algorithm's advantages and disadvantages are compared, can choose a good deal with 0/1 classification problem of logistic regression algorithm, and logistic regression algorithm to establish GPRS prediction model based on the loss of business. From Mobile Corporation in Guangdong Province GPRS service churn probability the top 5% and 10% users, the precision and recall rate and improve the three indexes of the model test, we found that the accuracy of the model and determine the user positioning loss loss of the user's coverage is high when compared, which can effectively predict the loss of customers using GPRS package. In addition, by comparing the logistic regression algorithm and use the model of decision tree algorithm application, results fully demonstrated that using logistic regression algorithm to build The GPRS business churn prediction model is more advantageous in practical applications. Finally, according to the effect of the model in solving the actual prediction problem, it further verifies that it has strong practicability.
【作者单位】：陕西省宝鸡教育学院数学系;西安工业大学机电工程学院;
【基金】：广东移动数据部流量业务专项运营项目(G001-YDSH-BX-140020)
【分类号】：TN929.532
【正文快照】： 通用分组无线服务技术(General Packet Ra-dio Service,GPRS),它是第二代全球移动通信系统向第三代全球移动通信系统的过渡、衔接性技术,GPRS业务是全球手机系统移动电话用户可用的一种移动数据业务[1],这项业务推动了电信事业的飞速发展.但是,随着各种实际因素不断增多,加之

【相似文献】