Statistical Method In Quantitative Trading

Along with the emergence of Index Future and short selling, the quantitative trading has been developing step by step in Chinese financial market. However, current financial market in China lacks diversity in trading targets and robustness in trading mechanism as well as participating investors are not so mature as those in western markets, which makes it impossible for the development of quantitative trading to proceed smoothly without a hitch. For instance, Everbright Securities’scandal happening at August16th last year was quite a failure.Once quantitative trading appears in western markets, statistical arbitrage has always been the emphasis of quantitative analysis research, whose major goals are to reduce systematic risk in market and to maintain the excess profits. Nevertheless, as the development of computer science, quantitative analysis research is no longer merely focusing at statistical arbitrage. Data mining techniques have already been applied into financial markets and delivered nice performance. We utilize the statistical methods in the liquidity problem, and analyze financial news items with the aid of latent semantic analysis techniques. Furthermore, the proceeded topics are thereafter analyzed in conjunctions with support vector machines algorithm to build a prediction model for stock index. The details are as follow:(1) Influence of Short-Selling Mechanism on Liquidity Based on the liquidity variation of500short-selling allowed stocks during the three years from2010to2013, we apply Wilcoxon rank-sum test to investigate whether short-selling adjustments have different effects on stocks with different market capitalization, return volatility, permission for short sales and listed markets. Furthermore, we employ trading data to identify the determinants of stock market liquidity, including total trading volume, permission for short sales, stock return volatility and whether listed on A market. Beside, the market liquidity in2013is significantly worse than2011. In conclusion, whether the market liquidity is able to be improved not only relies on expanding the range of short sales, but also depends on the enacted coordinated policies accepted by stock market participants.(2) Prediction for Stock Market based on Data Mining techniques In this paper, we attempt to explore the correlation between the topic distribution of financial news and the movement of stock market, and therefore deliver a prediction for stock market by analyzing the distribution. Specifically, we adopt probabilistic topic model with data mining techniques, and maintain the topic distribution by clustering financial news documents. The distribution is thereafter analyzed in conjunction with actual market data to understand its impact on the market, and is used for the prediction with aid of support vector machines algorithm. As shown in the experimental result, the topics about international commerce and utilization are strongly related with the movement of stock market, besides, an accurate prediction system is proposed and the trading strategy based on it delivers good performance.