Research On Persona Technolegy Based On Deep Learning

The user portrait is a technique which uses real data to paint a user's full picture.By establishing user portraits,companies can locate products more rationally.This technology has been widely used for its practicality and effectiveness.Besides,the rapidly developing deep learning technology gets increasing attention.How to combine user portrait with deep learning has become a new technical focus.This article focuses on how to use deep learning methods to paint user portraits more effectively.In this paper,we use a multi-layer perceptron and other related theoretical knowledge to design a scheme which paints user portraits effectively.Firstly,according to the user's URL access list,a tag system is constructed,and then the model is trained using the labeled data set,three models are designed for different characteristics of the data set.Then,the trained model is deployed on the distributed computing framework to complete the depiction of the user's portrait.The results show that tag models have different performance on data sets.For gender and age tags,the URL-based model has a better performance,and the accuracy is about 80%.For interest tags,the keyword model performs better,and the accuracy of some tags exceeds Seventy percent.Besides,by deploying on the Spark distributed computing framework,the model's running speed has increased by about 20 times.