摘要
在数据业务爆发式增长的时代,掌握海量用户数据的电信运营商占有价值优势。通过大数据技术对数据价值进行洞察识别和探索已经成为运营商发展增值业务的必要手段。对于传统的营销方式,各省市公司通过一些业务指标筛选目标客户或者单独建模的方式进行粗放营销,数据获取周期长,营销接触率不高。针对上述痛点,提出在小样本数据情况下构建相似群扩展服务(Lookalike)。该服务结合电信行业的数据及业务特点,集成了度量学习、深度学习等方法,有效支撑电信运营商实现精准营销。Lookalike服务减少了人工参与,自动实现营销目标客户的输出,大大提高了营销活动的工作效率及成功率,在多次实际电信项目中得到验证TTM(time to market)取得了由月到日的提升,营销成功率明显提高。
With the explosive development of data service,TSP（telecommunication service provider） has taken advantages of owing massive user data.It is a necessity for TSP to develop value added business using big data techniques which explores the data value.The traditional way of marketing utilizes rules or supervised classification method meets the challenge of low success rate and long data acquisition period.Therefore,a new service named Lookalike was proposed.The service supports precision marketing by integrating matric learning and deep learning efficiently based on telecom operators＇ data characteristics.The Lookalike service decreases the artificial participation and enhances the efficiency and success rate of marketing activity.The enhancement of TTM（time to market） has been improved in many real programs and the success rate of marketing has increased absolutely.
出处
《电信科学》
2018年第1期166-173,共8页
Telecommunications Science
