Online Advertising Click-Through Rate Prediction Based On Enhanced DeepFM Model With Feature Selection

With the fast development of recommendation system, search engine optimization, advertising and so on fields, click-through rate prediction has been extensively applied. Xgboost, Linear regression and GBDT are traditional machine learning methods which need a considerable amount of manual feature engineering process. When DNN model is widely used in estimating Click-through rate prediction in many studies like DeepFM, xDeepFM and Wide and Deep. For the clickthrough rate prediction task, we create an enhanced DeepFM model in this work. In addition, feature selection technique like Recursive Feature Elimination (RFE) is used to improve the prediction accuracy. The experiment is implemented on the Avazu CTR datasets with 10 days training set and 1 day test set. Experiments shows that the proposed model has better performance compared to the other models in terms of accuracy and auc score.