Mobile Application Recommendation Based On Content And User Interaction

With the rapid growth of mobile applications in major mobile app stores,it is difficult for users to choose their desired mobile applications.Therefore,it is necessary to provide a highquality mobile application recommendation mechanism to meet the user’s expectation.The existing mobile application recommendation methods have some problems,such as inaccurate text representation of mobile application,ignoring the original feature weight of mobile application and not considering the deep interaction between users and mobile application.Therefore,this paper proposes two popular mobile application recommendation methods at the content level and a personalized recommendation method at the user interaction level to solve the above problems.More specifically,the research contents of this paper are as follows:（1）In view of the inaccuracy of text representation,this paper proposes a mobile application recommendation method based on topic attention and factor decomposing machine.By introducing the topic attention mechanism,the LSA topic vector and Bi LSTM local semantic vector are used to calculate the weight coefficient of the hidden layer,so as to obtain better text vector representation and achieve better classification effect.On this basis,the query statements can be used to locate the most relevant APP categories,and then consider the interaction between features through factor decomposition machine model to improve the recommendation accuracy of APP.（2）As for the feature importance of mobile applications,this paper proposes a mobile application recommendation method based on feature importance.The model uses the SENET module to dynamically learn the importance of features,and uses more complex bilinear interaction layer to learn feature interaction.The final deep model can improve the accuracy of mobile application recommendation model.（3）Aiming at how to effectively code the deep interaction between users and mobile applications,this paper proposes a mobile application recommendation method based on user interaction.This method first constructs the interaction graph of user and app,and then mines the deep interaction relationship between user and APP by embedding propagation layer,which can further refine the representation of APP and user.Finally,the inner product is used to calculate users’ preferences for different APPs to complete the recommendation task.After our verification on the open dataset of Kaggle,the above three mobile application recommendation methods can achieve better experimental results compared with their respective mainstream baseline methods.