Research And Application Of Android Application Classification Method Based On Neural Network

With the continuous development of Internet technology and mobile smart devices,technology has continuously integrated into all corners of people's lives.The Android system has always occupied a place in the field of mobile intelligent terminals,and more and more Android applications have emerged at the historic moment.While enjoying the convenience brought by Android applications,users are also facing many problems.First,the emergence of many malwares poses a threat to the privacy and property security of Android users.In addition,if the correct function classification of the Android application is not performed,it is difficult for the user to find the Android application of the desired category in the massive Android application without knowing the name of the specific application.How to efficiently manage a large number of Android applications and build a good Android ecological environment is an important issue that needs to be studied and solved.This subject builds the Android application property classification model and Android application function classification model based on neural networks,and designs and implements the Android application classification system on this basis,which solves the classification problem of the property and function of Android applications.Aiming at the property classification of Android applications,this topic collected 6120 Android malicious applications and 5513 Android benign applications as samples of Android application property classification.This topic uniquely researched and verified the difference between the number of four components of the benign and malicious Android samples,and used it as one of the characteristics of the Android application property classification model,and constructed a neural network-based Android application property classification model,reaching 99.7%accuracy.Aiming at the function classification of Android applications,this subject conducts an in-depth study on Android application samples with different functions.This topic collected 13 kinds,a total of 5103 Android application installation packages and application introductions,as experimental samples of functional classification.This topic uniquely integrates the application introduction and string resources of the Android application into the description information of the sample,performs feature processing on the Android application sample based on natural language processing technology,builds a classification model based on the neural network,and improves the model performance through integrated learning,With an accuracy of 87.2%.In addition,this topic uniquely constructed the Android application description word cloud map to complete the visual auxiliary display of Android application function classification.On this basis,in order to meet the visual operation requirements of Android application classification prediction,this topic has developed an Android application classification system based on B/S architecture,which effectively provides Android application classification services with high visibility and friendly interaction.