基于CNN-GRU的移动APP流行度预测模型

Mobile APP popularity prediction model based on CNN-GRU

移动APP流行度预测对应用推荐、广告投放等意义重大。但是现有方法大多依赖手工特征工程,工作量大且效率较低。为此,提出一种基于深度神经网络的移动APP流行度预测模型。利用最大信息系数进行特征相关性分析以确保特征选取有效性,结合历史流行度特征,通过门控循环单元(gate recurrent unit,GRU)和注意力机制构建长期演化模型来推演发展趋势,基于多尺度卷积神经网络(convolutional neural networks,CNN)和注意力机制构建短期波动模型以实现预测动态优化,结合其他重要特征利用GRU和注意力机制建立多因素影响模型。通过时间注意力模块将上述模型融合,实现流行度预测。实验结果表明,所提模型在移动APP流行度预测方面相对更为精准有效。

Mobile app popularity prediction is crucial for app recommendation and advertising.However,existing methods rely heavily on manual feature engineering,which is labor-intensive and inefficient.This paper proposes a deep neural network model for mobile app popularity prediction.First,the maximal information coefficient is utilized to perform feature correlation analysis,ensuring the effectiveness of feature selection.Combined with historical popularity features,a long-term evolution model is constructed through a gate recurrent unit(GRU)and an attention mechanism to deduce the development trend.A short-term fluctuation model is constructed based on multi-scale convolutional neural networks(CNN)and an attention mechanism to achieve dynamic optimization of prediction.Also,other important features are incorporated to build multi-factor influence models using GRU and attention.Finally,the above models are integrated through a temporal attention module to achieve popularity prediction.Experimental results demonstrate the relative accuracy and effectiveness of the proposed model in predicting the popularity of mobile apps.