Research On Stock Matching Trading Strategy Based On Radial Basis Neural Network Algorithm

Quantitative investment is a rational investment method and plays an increasingly important role in the financial market.Quantitative investment can overcome human's greed and fear to the greatest extent,gently fluctuate the income,and quantitative trading has the advantages of easy expansion of scale and fast operation,which is conducive to improve work efficiency and accelerate decision-making.As a neutral trading strategy of quantitative trading,paired trading shows excellent profitability and is increasingly favored by investors.The machine learning algorithm has the advantages of intelligent decision-making and pattern recognition,among which the radial basis neural network has the advantages of high training efficiency and avoiding local optimization.Therefore,it is of great significance to use RBF neural network to optimize the pairing-based trading strategy.In this paper,the historical stock price data of margin and short selling stocks in domestic a-share market are taken as the research object.Through the analysis of cointegration relationship,the optimal stock pairs with stable co-integration relationship are selected.Stock on the share price spread sequence is then used as the raw data training RBF neural network,using the RBF neural network to matching threshold,the open positions threshold,stop threshold is optimized,so as to achieve the aim of improving pairs trading strategy profitability,and then through the shares of real data to test the optimized strategy deals to test the effectiveness of the optimized matching trading strategies.The study draws the following main conclusions :(1)for open trade,if the spread reaches the open line,and if the radial basis neural network makes the prediction that the spread will continue to expand,the open position will be delayed to obtain the maximum profit of a single trade;(2)for stop loss trading,if the spread reaches the stop loss line,and if the radial basis neural network makes a prediction that the spread will change from expansion to reduction,the stop loss trading will be rejected to avoid wrong stop loss;(3)the results of the strategy test showed that the asset income increased by 9.6%,the sharpe ratio increased by 0.09,the information ratio increased by 0.987,and the maximum withdrawal decreased by 12.21%.