Research And Application Of Algorithmic Trading Using Large-Scale Machine Learning

As the development of computer technology, more and more data is collected and stored. Statistics rules that were found in big data could be used for the investment guide. Algorithmic trading is expected to find investment opportunities from the big data through mathematics and computer models, auto-complete investment in securities trading. Compared to the fundamental analysis and technical analysis, algorithmic trading has advantages of objective, disciplined and accurate.Algorithmic trading has existed for many years, and plays an important role in the stock market, in foreign countries. As constantly enrich of the transaction variety in Chinese stock market recent years, algorithmic trading is increasingly valued and tried in kinds of investment institutions. While its system development is extremely urgent, with the learning of Everbright security issue in the summer of2013.Based on the demand analysis, this paper makes a system design in detail, which will be divided the algorithm trading system into three modules, data collection and storage module, quantitative models runtime module and programmatic execution module.Data capture and storage module is responsible for obtaining data from different data sources and storage there. Machine learning model running on the quantitative model runtime module, try to find investment opportunities from those data, then place order to buy and sell securities.The existing algorithm trading system is designed to mainly use for processing structured data obtained from the stable data sources, such as transaction records. In the face of large-scale and complicated internet data, the existing algorithm trading system is less to use due to the existence of data capture and storage difficulty.The theme crawler technology has been researched widely in vertical search, mass data storage problem as an important part of big data technologies, also has been widely discussed.The theme crawler and mass data storage technology are applied to the algorithmic trading system in this paper. The topic crawler used SVM technology to improve the efficiency of capturing data. Sharding technology ensured the storage system support transaction and deal with big data storage. At the same time, taking the distributed database as the core, the trading system can quickly adapt to different data sources.In addition to the realization of data capture and storage module, this paper also built quantitative model runtime modules, and made the application of SVM in financial time series. Combined with transaction experience, this paper constructs an intraday trade model, and the prediction accuracy is slightly higher than that of the existing research. All those contribute to show whole structure of an algorithmic trading system.