Research On Financial Text Classification Method Based On Deep Learning

Financial market has been playing an important role in modern social economy,and there is an interaction between financial news and financial market.Classified financial news texts can help financial individuals or financial institutions master more detailed news in a certain sub category,so as to make correct decisions.Especially for professional financial experts,after classification,the detailed and effective financial texts can grasp the current advanced research technologies and the possible future research directions,and can comprehensively understand the financial information in the network.At present,text classification is mostly studied based on the existing public data set,but rarely targeted at text classification in the professional field,which leads to the lack of financial professional text classification methods.In order to solve the above problems,this paper takes the financial news as the data set,and uses the deep learning method to classify the financial news texts.The main work and innovation of this paper are as follows:(1)Construct the text data set of Chinese financial news.In view of the fact that there is no publicly available Chinese financial news text data set on the Internet,this study uses crawler technology to extract tens of thousands of texts from authoritative Chinese financial websites as the data set.The data set in the financial field lays a good foundation for the deep learning of financial text classification model.(2)Systematically propose the indicators to verify the availability of data set for the first time.The quality of data set can directly affect the performance of deep learning,and there is no systematic and effective verification method at this stage.Therefore,this paper presents the indicators to verify the usability of data set for the first time.The specific indicators include: the rate of data redundancy,the rate of data noise,the rate of category imbalance,category and content of the matching degree,composition deviation,etc.The purpose is to verify whether the constructed data set is available through these indicators,and to promote the development of deep learning from the perspective of data set.(3)The deep learning method based on character level is used to classify Chinese financial news texts for the first time.Compared with English text classification,Chinese text classification is more difficult due to the complexity of Chinese.At the same time,the text classification in this study is based on the classification of subclasses in the same professional domain data set,and the differences between subclasses in the same domain are less than the differences between different domain categories.Previous researches are mostly based on the text classification under various categories,but few subclasses based on the same field.Therefore,this paper proposes an ADchar CGNN neural network model for the classification of Chinese financial texts.The experimental results show that the accuracy of the network for Chinese financial texts can reach96.45%.The network,based on character levels,also makes it easier to handle financial text that may contain Chinese,English,numbers,and other types of characters.This paper constructs the Chinese financial news text data set,puts forward the evaluation of the data set,and constructs the financial text classification model based on parts.In the future,there is a possibility of further improvement in the specific application and practice.In the future,a cross-language text classification model can be built,so that the model can be applied not only to Chinese,but also to other languages such as English.At the same time,the network structure proposed in this paper has a deep level and cannot achieve the ideal effect for small data sets.Therefore,the model can be improved in the future so that small data sets can also participate in it.