面向量化交易的金融数据处理平台研究与原型实现

发布时间：2018-06-28 22:26

本文选题：量化交易 + 金融数据处理工具　；参考：《电子科技大学》2016年硕士论文

【摘要】：随着社会经济和互联网的发展,中国金融市场迎来了空前的繁荣,投资者在面对机遇的同时,也伴随着诸多的挑战和风险。千万级的股民数量,2500多支A股以及便捷的股票交易方式使得每个交易日产生的交易数据都很庞大,而且每日与金融相关的新闻资讯也在网络上不断更新,股民使用原始的投资手段已经难以应付如此大量的数据信息。因此,结合了数理统计和计算机技术的量化交易对解决上述问题具有重要的意义和价值,引起了研究人员广泛的关注。本文构建了一个提供给投资者进行算法研究的面向量化交易的金融数据处理平台,旨在提供给用户有效的金融数据处理工具,并在该平台基础之上实现了三个可行的策略思路供算法研究员参考。本文首先对量化交易的国内外发展现状进行了总结,并对现有的量化交易手段进行了研究和分析,研究了时间序列分析法以及文本分析法在量化交易中的应用。然后针对金融数据处理平台的实际需求进行了详细的需求分析,据此设计出了系统的整体架构,分解出了辅助功能模块和算法实体模块,分别对其关键功能进行详细的设计。最后实现了具有多个工具的金融数据处理平台,本文的工作量主要在以下四个方面:(1)在辅助功能模块中实现了四种金融数据处理工具。一是使用网络爬虫获取金融相关的新闻数据;二是使用Node.js的Addons技术改进了股票交易数据获取平台;三是实现了能在任何历史时刻进行交易模拟的程序;四是基于eCharts.js技术实现了平台的可视化分析与呈现。(2)在基于文本处理的策略模块里提出了基于TF-IDF的朴素贝叶斯模型的新闻情感倾向预测;并使用情感词典来量化股评,将结果作为量化择时特征数据的一部分。(3)在量化选股模块里提出了基于多项式线性回归模型的多因子策略来实现量化选股,该策略根据历史股票交易数据、基本面数据以及衡量系统风险的?值来推荐股票组合。(4)在量化择时模块里,本文从情绪指标、市场前期走势、经济指标、货币环境这四个方面提取出数据特征,并使用支持向量机作为训练模型。
[Abstract]:With the development of social economy and Internet, China's financial market is facing unprecedented prosperity. Investors face opportunities, but also with a lot of challenges and risks. With more than 2,500 A-shares and convenient stock trading methods, the number of investors at the level of 10 million makes the transaction data generated on each trading day very large, and daily news information related to finance is constantly updated on the Internet. It is difficult for investors to deal with such a large amount of data by using primitive investment methods. Therefore, the combination of mathematical statistics and computer technology is of great significance and value in solving the above problems, and has attracted extensive attention of researchers. In this paper, we construct a financial data processing platform for quantitative transactions, which is designed to provide users with effective financial data processing tools. On the basis of the platform, three feasible strategic ideas are implemented for the reference of the algorithm researcher. Firstly, this paper summarizes the development of quantitative trading at home and abroad, studies and analyzes the existing quantitative trading methods, and studies the application of time series analysis and text analysis in quantitative transactions. Based on the detailed requirement analysis of the financial data processing platform, the overall architecture of the system is designed, the auxiliary function module and the algorithm entity module are decomposed, and the key functions are designed in detail. Finally, a financial data processing platform with multiple tools is implemented. The workload of this paper is mainly in the following four aspects: (1) four kinds of financial data processing tools are implemented in the auxiliary function module. First, using web crawler to obtain financial news data; second, using Node.js Addons technology to improve the stock trading data acquisition platform; third, realizing the procedure of trading simulation at any historical time; Fourthly, it realizes the visual analysis and presentation of the platform based on eCharts.js technology. (2) in the strategy module based on text processing, the prediction of news emotion tendency based on the naive Bayesian model of TF-IDF is put forward, and the emotion dictionary is used to quantify the stock review. The results are taken as part of the quantitative timing feature data. (3) in the quantitative stock selection module, a multi-factor strategy based on polynomial linear regression model is proposed to realize quantitative stock selection, which is based on the historical stock trading data. Fundamental data and measures of systemic risk? (4) in the quantitative timing module, this paper extracts the data features from four aspects: emotion index, early market trend, economic index and monetary environment, and uses support vector machine as the training model.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.52

【参考文献】