Research On User Profiling Construction

With the rapid development of Internet technology,people’s daily life is more and more inseparable from the Internet and mobile Internet,the Internet has long been an important channel to access to information(such as online shopping,news,etc.)However,the information people face in the Internet is exponential growth,and most of them are useless information.It has become an urgently need to be solved problem that quickly extract the different characteristics of user groups and construct user profiling to meet the individual needs of different users.In this paper,we researched two different areas of user profiling construction.(1)Research on user profiling construction for the state grid usersThe State Grid users who are sensitive to electric charge often have a strong reaction on electric quantity,electric price,electric charge,payment,arrearage and other electrical service caused by electricity consumption.How to rapidly locate the electric-charge-users has an important role in reducing customer complaints rate,enhancing customer satisfaction,and establishing a good service image of power supply enterprise.Based on the grid users data,this paper presents a multi-view ensemble framework for constructing user profile,to quickly and accurately identify the electric-charge-users.First of all,this paper analyzes the grid users and proposes to use two channels to model the users with different characteristics respectively.Secondly,this paper presents a variety of feature extraction methods for constructing user multisource feature systems.Finally,in order to make full use of multi-source features,this paper proposes a multi-view ensemble model.With this method,we managed to obtain the F1 score of 0.90379(The first place)in the “User Profile” contest of 2016 CCF Big Data and Computational Intelligence Contest,the results suggest that method is effective.(2)Research on user profiling construction for the Sina weibo usersAs Sina weibo is the largest microblogging social network in China,a large number of users browse and publish microblog on it every day.In the face of massive information,the weibo user profiling construction hold an important part in public opinion monitoring,advertising accurate delivery,social conditions and people’s livelihood reflection and so on.This paper studies the classification of users’ age in weibo user profiling.First of all,we construct a multi-granularity feature system.According to the different types of weibo features,the features are divided into textual features and social features.Then we divide the textual features into weibo-granularity and user-granularity according to the different characterization angle.And based on the convolution neural network,this paper proposed a neural network which can combines the multi-granularity features,so that we can make full use of the features obtained by the previous stage.And due to the different models for the use of the features of different degrees,so based on the neural network we combine a variety of machine learning models for extracting textual features and blend multi-model results as the final output.With this method,we managed to obtain a good result on the age classification task of 2016 SMP CUP Weibo User Profiling Contest.