基于聚类分析算法的野呼用户识别模型研究

Research on identification model of wild call user based on cluster analysis algorithm

针对近年来由野呼引起的用户投诉数量不断上涨,严重影响运营商品牌形象等问题,本文提出一种基于聚类分析的野呼用户识别新算法。该算法对运营商数据重新进行编码及预处理,基于K-means算法对训练样本进行无监督建模,把样本自动分为两类,找到两类的分界线,即为野呼用户筛选规则阈值。最终把该模型实施于智能管理平台,通过对多个月野呼投诉率统计,野呼投诉工单占比下降5.5个百分点,大大提升了客户满意度,维护运营商形象,降低运营成本。

In response to the increasing number of user complaints caused by"wild calling"in recent years,which seriously affects the brand image of operators,a new algorithm for identifying users of wild calling based on cluster analysis is proposed.The method firstly re-encodes and preprocesses the operator data,uses the K-means clustering algorithm to perform unsupervised modeling of the training samples,automatically divides the samples into two categories,and finds the boundary between the two categories,which is called wild call.User fi ltering rule threshold.Finally,the model was implemented on the intelligent management platform.Through the statistics of the complaint rate of out-of-call calls for multiple months,the complaint rate was reduced by nearly 8 times,which greatly reduced the customer complaint rate,maintained the image of the operator,and reduced operating costs.