摘要
本文使用某电商公司提供的广告点击日志流,构建基于用户画像的互联网广告点击率预测xDeepFM模型。研究发现:xDeepFM模型对预估准确率提升明显;用户画像系统可以很好的解决深度模型对高维稀疏特征的学习问题,有效提高预估准确率。
This paper uses the click log stream provided by an e-commerce company to build the x DeepFM model of Internet advertising click rate prediction based on user portrait.The results show that:x Deep FM model can significantly improve the prediction accuracy;user portrait system can solve the problem of learning high-dimensional sparse features from depth model,and effectively improve the prediction accuracy.
作者
周亲
吴运辰
吴俊坤
ZHOU Qin;WU Yunchen;WU Junkun(School of economics,Nanjing University of Posts and telecommunications,Nanjing Jiangsu 210023)
出处
《软件》
2021年第2期171-174,共4页
Software