A Network Based On Deep Interest Network for Taobao Click-Through Rate Prediction
Fang Yingyi, Zhenghan Chen, Yunxiao Ma, Lanchuan Lin · 2021
Improving the prediction accuracy of advertising click-through rate (CTR) is a very important task in the field of computing advertising. The higher the accuracy of click-through rate prediction, the more accurate the information provided to users will be, and the better the promotion effect of merchants will be, which will increase the profits of media platforms, DSP companies and merchants. In this paper, we propose a new model based on DIN (Deep Interest Network), which can make better use of data structure and achieve diversity and local interest to users through a network structure similar to attention. activation. Facts have proved that this method is effective and significantly better than traditional models. Experiments show that the model can bring more interpretability and obtain good AUC (area under the curve) performance.