Deep Interest Network (DIN) is a model designed for click-through rate (CTR) prediction in display advertising. Traditional methods use fixed-length vectors to represent user interests, which limits their ability to capture diverse user interests. DIN addresses this by introducing a local activation unit that adaptively learns user interest representations based on historical behaviors relevant to a specific ad. This allows the model to generate varying representations for different ads, enhancing its expressive ability. Additionally, DIN incorporates two techniques: mini-batch aware regularization and data adaptive activation functions, which improve training efficiency and performance for large-scale deep networks. Experiments on public datasets and Alibaba's real-world data show that DIN outperforms state-of-the-art methods in CTR prediction. DIN has been successfully deployed in Alibaba's online display advertising system, significantly improving business performance. The model's ability to adaptively capture user interests through local activation mechanisms and its effective training techniques make it a powerful solution for CTR prediction in industrial applications.Deep Interest Network (DIN) is a model designed for click-through rate (CTR) prediction in display advertising. Traditional methods use fixed-length vectors to represent user interests, which limits their ability to capture diverse user interests. DIN addresses this by introducing a local activation unit that adaptively learns user interest representations based on historical behaviors relevant to a specific ad. This allows the model to generate varying representations for different ads, enhancing its expressive ability. Additionally, DIN incorporates two techniques: mini-batch aware regularization and data adaptive activation functions, which improve training efficiency and performance for large-scale deep networks. Experiments on public datasets and Alibaba's real-world data show that DIN outperforms state-of-the-art methods in CTR prediction. DIN has been successfully deployed in Alibaba's online display advertising system, significantly improving business performance. The model's ability to adaptively capture user interests through local activation mechanisms and its effective training techniques make it a powerful solution for CTR prediction in industrial applications.