MR-FI: Mobile Application Recommendation Based on Feature Importance and Bilinear Feature Interaction

With the rapid growth of mobile applications in major mobile app stores, it is challenging for users to choose their desired mobile applications. Therefore, it is necessary to provide a high-quality mobile application recommendation mechanism to meet the user’s expectation. Although the existing methods make significant results on mobile application recommendation, the recommendation accuracy can be further improved. More exactly, they mainly focus on how to better interact between mobile applications’ features, but ignore the importance or weight of these features themselves. Based on squeeze-excitation network mechanism and bilinear function with combining inner product and Hadamard product, this paper proposes a mobile application recommendation method based on feature importance and bilinear feature interaction to solve this problem. First of all, it exploits a SENET (Squeeze-Excitation Network) mechanism to dynamically learn the importance of mobile applications’ features and uses a bilinear function with combining inner product and Hadamard product to effectively learn these features interactions, respectively. Then, the user preferences for different mobile applications are predicted through infusing cross-combined features into a deep model via integrating the classic deep neural network component with the shallow model. The real dataset of Kaggle is used to evaluate the proposed method and the experimental results show that the method can achieve the best results in most cases in terms of AUC and Logloss. It can effectively improve the recommendation accuracy of mobile applications.