一种改进的基于通话行为特征的骚扰电话检测方法
2021-01-28张瑾
张瑾

摘要:移动通信的快速发展为人们带来了便利,同时也伴随着商业目的的营销甚至一些非法信息的传播,导致骚扰电话的泛滥。骚扰电话已成为用户投诉电信运营商的重要原因之一,也对通信设备的负荷造成了重大影响。本文通过对用户通话行为特征进行分析,从运营商数据中提取了包含传统特征的多维融合特征,如平均通话距离、外卖时间占比、物理位置变化率等生成骚扰电话识别模型,有效地识别了骚扰电话,同时降低了快递外卖以及一些社区医院、水站等符合社区居民利益的正常电话的误检测率,并通过投诉数据和实际检测结果不断优化已有模型,提高识别准确率。
关键词:骚扰电话;运营商数据;通话行为特征
Abstract: The rapid development of mobile communications has brought convenience to people, accompanied by marketing for commercial purposes and even the spread of some illegal information. The proliferation of harassing calls has become one of the important reasons for users to complain to telecom operators, and has also caused a significant impact on the load of communication equipment.This article analyzes user call behavior features and extracts multi-dimensional fusion features including traditional features from operator data, such as average call distance, take-out time ratio, and physical location change rate to generate harassing phone recognition models. It reduces the false detection rate of delivery and some normal calls such as community hospitals and water stations, and optimizes existing strategies through complaint data.
Keywords: harassing calls; mobile operator data; call behavioral feature
1 引言
随着移动通信业务不断发展,移动通信在给人们带来了便利的同时,也使一些人为了商业目的而进行电话营销、骚扰。其中骚扰电话的泛滥已成为用户投诉电信运营商的重要原因之一,也对通信网络设备的负荷造成了重大影响[2]。2018年7月30日,工信部等13个部门联合发布《综合整治骚扰电话专项行动方案》的通知,通知里提出,即将在全国范围内严厉整治骚扰电话乱象。
本文首先通过对运营商信令数据和话单数据进行分析,得出典型场景下的骚扰电话的通话行为特征。然后通过大数据技术[4],对运营商的信令监测数据进行计算,得出每个用户的通话行为特征值,生成骚扰电话识别模型,经过识别模型的过滤,得出疑似骚扰电话号码。……
