Internet website automatic classification method based on deep learning

Abstract

The present invention relates to an Internet website automatic classification method based on deep learning. The method comprises the steps of: collecting a lot of original description information ofan internet website according to a DNS server log as a website data set, and performing pre-processing and manual tagging; extracting high-dimensional feature vector representation, configured to input a deep learning model, of each website, increasing corresponding website category tags for each website, and converting the corresponding website category tags to type vectors; taking high-dimensional feature vectors as input of the deep learning model, taking the category vectors as output of the deep learning model, and employing an Adam gradient descent algorithm optimizer to supervise and train a recurrent neural network deep learning model based on the LSTM; increasing one layer of SoftMax regression behind the trained LSTM recurrent neural network deep learning model; and taking the website category corresponding to the dimension of the maximum probability value in the probability distribution vector as a website category, and comparing the output network category with the actual category of the website to obtain a classification accuracy of an Internet website.

Description

technical field [0001] The invention relates to a method for automatically classifying Internet websites based on deep learning, which belongs to the technical field of artificial intelligence. Background technique [0002] In recent years, the Internet has grown explosively, and China's development speed in the Internet field has amazed the world. The 40th "Statistical Report on Internet Development in China" issued by China Internet Network Information Center in July 2017 [1] It pointed out that as of June 2017, the number of Chinese netizens reached 751 million, and a total of 19.92 million new netizens were added in half a year. The Internet penetration rate has reached 54.3%, while the total number of Chinese websites has reached 5.06 million, and the number of websites under the ".CN" domain name has reached 2.7 million. my country's Internet has shown an upward trend in terms of the number of Internet users, Internet penetration rate, and the number of websites. [...

AI Technical Summary

Problems solved by technology

[0015] In short, the shortcomings and deficiencies of the existing technology are: the amount of data used for model training is not large enough, not new enough, the amount of information based on classification is insufficient, and the methods and models used are not suitable for classifying high-dimensional data samples, extracting features and learning information Insufficient ability, the number of categories is not enough, it is difficult to meet the needs of multi-category

Images