Research And Prototype Implementation Of A Quantitative Transaction Oriented Financial Data Processing Platform

With the development of social economy and the Internet, Chinese financial markets ushered in an unprecedented prosperity, which means that investors can get more opportunities but accompany with many challenges and risks at the same time. In every trading day, the trading data turns so huge because of tens of millions of investors, more than 2500 A-shares and convenient ways of stock trading. And the financial news and information updated on the network constantly are updated on the network. It is difficult for the investors to cope with such a large amount of data information by original means of stock investment. To solve the above problems, the quantitative trading which combined with the mathematical statistics and computer technology will have great significance and value.This thesis describes a financial data processing platform designed for the investors who are researching quantitative trading, which provides the user convenient financial data processing tools. It also realizes three quantitative trading strategy models on the basis of financial data processing tools. First of all, it summarizes the domestic and international development present situation of quantitative trading and analyses the existing methods of quantitative trading. The method of time series analysis and text analysis on the application of quantitative trading were studied. Then, the demand analysis was analyzed based on the actual demand of the financial data processing platform. Hereby, we designed the overall system architecture and decomposed the assistant function module and algorithm entity module. In order to make the platform better, it also designs the key function in detail. Finally, it realizes convenient tools for the financial data processing platform. The main contents are as follows:(1) In the auxiliary function module, it realizes four convenient financial data processing tools. One of tools is using a web crawler to obtain financial related news data and another is using the Node.js’ s Addons technology to improve the stock transaction data acquisition platform. The third is the realization of the simulated trading process at any moment in history. The last one is based on eCharts.js technology, completed the platform of visual analysis and rendering.(2) In the investors mood prediction module, it proposes prediction based on TF-IDF naive Bayes model News emotional tendency. This strategy is to predict investors mood according to the relevant financial data of network news and stock transaction data and it also bases on sentiment dictionary investors mood quantitative value as part of quantitative timing characteristics of data.(3) In the quantitative stock module, it proposes a strategy based on linear and polynomial regression model of factor strategy to achieve quantitative stock. This strategy is to select the relative quality of the stock portfolio, according to the historical transaction data of stocks, the fundamentals of data and the values of systematic risk.(4) In the quantitative timing module, this paper extracts the features of the data from the four aspects of the investor indicators, the market trend, the economic indicators and the monetary environment, and uses the support vector machine as the training model.