arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.05312cs.LGcs.IRstat.ML
收藏

基于特征交互的神经网络点击率预测

Feature Interaction based Neural Network for Click-Through Rate Prediction

Dafang Zou, Leiming Zhang, Jiafa Mao, Weiguo Sheng

摘要:

点击率预测是广告计算和推荐系统中最重要和最具挑战性的问题之一。为了使用这些数据构建机器学习系统,正确地对特征之间的交互进行建模是很重要的。然而,目前的许多工作中,计算功能的相互作用,在一个简单的方式,如内积和元素的产品。本文旨在充分利用特征之间的信息,提高深度神经网络在CTR预测任务中的性能。本文提出了一种基于特征交互的神经网络(FINN),它能够通过三维关系张量来建模特征交互。FINN提供了底层特征交互的表示和神经网络在建模高阶特征交互时的非线性。与经典基线相比,我们评估了我们的模型在CTR预测任务上的表现,并表明我们的深度FINN模型优于其他最先进的深度模型,如PNN和DeepFM。评估结果表明,特征交互包含更好的CTR预测的重要信息。这也表明我们的模型可以有效地学习特征交互,并在现实世界的数据集上取得更好的性能。

英文摘要:

Click-Through Rate (CTR) prediction is one of the most important and challenging in calculating advertisements and recommendation systems. To build a machine learning system with these data, it is important to properly model the interaction among features. However, many current works calculate the feature interactions in a simple way such as inner product and element-wise product. This paper aims to fully utilize the information between features and improve the performance of deep neural networks in the CTR prediction task. In this paper, we propose a Feature Interaction based Neural Network (FINN) which is able to model feature interaction via a 3-dimention relation tensor. FINN provides representations for the feature interactions on the the bottom layer and the non-linearity of neural network in modelling higher-order feature interactions. We evaluate our models on CTR prediction tasks compared with classical baselines and show that our deep FINN model outperforms other state-of-the-art deep models such as PNN and DeepFM. Evaluation results demonstrate that feature interaction contains significant information for better CTR prediction. It also indicates that our models can effectively learn the feature interactions, and achieve better performances in real-world datasets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑