论文部分内容阅读
为提高广告点击率的预估准确率,从而提高在线广告的收益,对广告数据进行特征提取和特征降维,采用一种基于LSTM的改进的递归神经网络作为广告点击率预估模型。分别采用随机梯度下降法和交叉熵函数作为预估模型的优化算法和目标函数。实验表明,与逻辑回归、BP神经网络和递归神经网络相比,基于LSTM改进的递归神经网络模型,能有效提高广告点击率的预估准确率。该模型不仅有助于广告服务商制定合理的价格策略,也有助于广告主合理投放广告,实现广告产业链中各个角色的收益最大化。
In order to improve the prediction accuracy of the advertisement click rate, the benefit of the online advertisement is improved, and the feature extraction and feature dimension reduction of the advertisement data are adopted. An improved recurrent neural network based on LSTM is used as the prediction model of the advertisement click rate. The stochastic gradient descent method and cross entropy function are respectively used as the optimization algorithm and the objective function of the predictive model. Experiments show that, compared with logistic regression, BP neural network and recurrent neural network, the improved recurrent neural network model based on LSTM can effectively improve the estimated accuracy of ad click rate. This model can not only help advertisers to formulate reasonable price strategies, but also help advertisers to put advertisements appropriately and maximize the benefits of various roles in the advertising industry chain.