Abstract
Click-through rate (CTR) prediction is an important issue in online advertising and recommendation systems. It is used to estimate the likelihood that a user will click on ads. The method used in traditional CTR prediction tasks is nothing more than the combination of feature engineering +LR/FM. This kind of work to improve the prediction effect through a large number of feature engineering is timeconsuming and laborious, and the construction process is not universal. Due to the sparse and highdimensional characteristics of data features, it is necessary to measure the importance of sparse features and obtain efficient feature interactions. A novel CTR model based on Mutual Information and Feature Interaction (MiFiNN) is proposed to measure feature importance and fine-grained feature interaction. To weigh the importance of sparse features, we use mutual information to find the correlation between each sparse feature and click behavior, and assign weights to features based on their contribution to the click results. The weighted features are interacted by combining the outer product with the inner product. The resulting feature interactions and the dense features set of the original input are taken as DNN inputs. We verify the proposed model on four datasets. In addition, several widely known models are introduced for comparison. The experimental results indicate the superiority of the model.
Funding Information
  • National Science Foundation of China (61472095)
  • Natural Science Foundation of Heilongjiang Province (LH2020F023)