通信与信息技术Issue(z1)：31-34,4.
基于PSO-LightGBM的用户订购预测-探索IPTV营销增长路径
User subscription prediction based on PSO-LightGBM--Exploring the growth path of IPTV marketing
DENG Zhiqiang 1YU Zhenyang 1YANG Lin1
作者信息
- 1. China Telecom Corporation Limited Sichuan Branch,Chengdu 610000,China
- 折叠
摘要
Abstract
With the deep integration of smart TV services and Internet Protocol Television(IPTV),accurately predicting users'subscription behaviors has become an important means of improving operational efficiency and boosting business revenue.Based on nearly one billion live and on-demand viewing logs from millions of users on the Sichuan Telecom IPTV platform,this study proposes a Particle Swarm Optimization-enhanced LightGBM model(PSO-LightGBM)to predict subscription probabilities for a specific monthly video package product using data mining and machine learning techniques.A total of 52 user viewing features were extracted,covering overall behavior,content preferences,temporal pat-terns,and interactions with popular content.Positive and negative sample sets were constructed for supervised learning.Experimental results show that the PSO-LightGBM model outperforms traditional machine learning methods across multiple evaluation metrics,achieving a prediction accuracy above 0.9.During a 14-day online validation period,the model maintained an average recall rate of 0.67 and increased the daily sub-scription volume by 12.33%,demonstrating its effectiveness and practical value.This approach provides strong support for personalized recom-mendation and precision marketing on IPTV platforms.关键词
IPTV/机器学习/LightGBM/粒子群优化Key words
IPTV/Machine learning/LightGBM/Particle swarm optimization分类
信息技术与安全科学引用本文复制引用
DENG Zhiqiang,YU Zhenyang,YANG Lin..基于PSO-LightGBM的用户订购预测-探索IPTV营销增长路径[J].通信与信息技术,2025,(z1):31-34,4.
