Construction Of Stock Classification System Based On Time Series Clustering And Complex Network

In recent years,artificial intelligence technology has been gradually integrated into people’s lives,providing convenience for human life.At the same time,economic growth has promoted the accumulation of wealth,and intelligent investment has attracted more and more people’s attention.People hope to assist people in securities investment through artificial intelligence.Compared with successful artificial intelligence cases,such as Alpha Go,intelligent securities investment is a game played by many people with complex rules and incomplete information,which is the development of artificial intelligence from a simple application scenario to a higher level.The most important thing for people to invest in stocks is to screen out suitable stocks.With the rapid development of China’s securities market,there are a large number of stocks.It is necessary to classify the stock market to improve investment efficiency.But at present,the market mainly divides stocks according to the artificial way,which has time lag problem and cannot fully express the dynamic changes of the market.Therefore,artificial intelligence methods will be used to divide the stock market,so as to achieve not only static description of the correlation between stocks,but also dynamic expression of the stock market.The research contents of this paper can be separated into the following three aspects:Construct the stock knowledge base,mainly including the stock price and attribute data set.Data is the basis for follow-up research.This article mainly draws data from official websites,some mainstream financial service websites,and professional financial data packages.The purpose is to verify the data to ensure the accuracy and integrity of the data.At the same time,the quantitative attributes of stocks are collected and filtered,in order to obtain the quantitative factors.The quantitative categories of stocks is generated by the quantitative factors.The classification is grouped by clustering based on the time series data of the stocks.On the basis of the dynamic time warping algorithm,the soft alignment technique is introduced to calculate the similarity of the two time series.The classification of the stock market is carried out in combination with the affinity propagation algorithm,so as to solve the shortcoming that the stock categories cannot be dynamically adjusted according to the market conditions in practice.The experimental results show that the improved DTW algorithm has a better effect on the similarity calculation of time series.The complex network theory is used to generate the stock network,and the improved Fast unfolding community discovery algorithm is used to carry out the group discovery in the stock network,in order to generate stock categories.This article expands on the basis of the traditional five categories of industry,region,concept,index,and user-defined,add quantitative classes,time series clustering and user classes to build a more complete stock classification system.Thus,the scope of application of user preferences has been significantly improved.At the same time,a visual display system is realized to present the entire classification system.The system helps users have a better understanding of the market situation through the top-down and bottom-up methods,thereby improving the efficiency of information acquisition.