Study On The Technology Of Portrait Of Natural Person Taxpayer

In the era of big data,social networks and social network-based Internet applications have been developing rapidly,and the explosion of data has been behind them.The collection and reasonable recommendation of user data and the effective storage and calculation of user portraits by using big data technology have gradually become a research hotspot.In the process of paying taxes,taxpayers produce a lot of data.By analyzing and collating these data,we can find out the tax payment process and habits of taxpayers,and find out a good way to improve tax service.Generally speaking,taxpayers have to go to the bank to pay their own taxes,sometimes causing unnecessary trouble because they miss the time to pay their taxes.This thesis designs a software installed on the App,which can timely inform taxpayers to deal with tax.At the same time,this software can also collect users' interests,hobbies and daily activities and other information.In the process of using the software,taxpayers can obtain the content services they are interested in,or remind them to pay tax in their spare time according to their daily activities.In order to provide users with better services,this topic takes user preferences as an example.By collecting user information,a subsystem can be realized to automatically provide personalized content push for users.Based on the research of portrait modeling technology and application,this thesis realizes the architecture of personalized push technology and summarizes the push process of personalized technology.By installing mobile App software,the system automatically collects user data.These data are classified into static and dynamic data,the basic information of users is static data,and the behavior data is dynamic data.According to these two different data information,a tag library is established.On the basis of the tag system,a day is divided into eight different time periods according to People's Daily activity rules,and then the number of users' interest labels in each time period is calculated.In data mining,the method and framework of user behavior data collection are designed to collect and calculate location information and intelligent terminal data more accurately.The degree of preference of the user to the label is analyzed to serve as the personality portrait of the user.By analyzing mobile users' behavior and hobbies,a classification algorithm is proposed to predict users' interests of different ages and genders.In this thesis,three classification algorithms are used,namely SVM(Support Vector Machine),BP(Backpropagation)neural network and DNN algorithm.This thesis designs and verifies the algorithm of personalized information push.Through the theoretical research of the three prediction algorithms and the comparison with the experimental research on data collection,the accuracy and reliability of the three prediction algorithms are obtained.Finally,it is decided to choose DNN(Deep Natural Network)algorithm with high prediction accuracy of the experimental results as the prediction algorithm of the subject.Finally,according to the user's location and time,combined with the user's portrait and interest tag,personalized push to the user based on DNN algorithm.The experimental results show that the personalized push system of DNN algorithm can provide personalized content to users' mobile phones according to their location and time and situation changes.Compared with traditional push,personalized content has better performance and is more popular.