基于概率神经网络的期货交易趋势识别及交易系统信号的优化

The Future Trading Trend Identification and Trading System Signal Optimization Based on Probabilistic Neural Networks

无论是技术面还是基本面分析,都是对于市场走势的一种预测。这种预测的基础就是以原有的已经发生过的市场条件作对比,当与以往交易条件重合率达到一定程度的时候,就会认为同样的趋势会再次发生。基于概率神经网络,选取数据为大连交易所豆粕主连的日线交易数据,采用小周期交易趋势(即5交易日数据)为交易趋势基础形态,数量化各交易形态,分为9种主要交易形态,利用PNN网络进行分类识别,判断趋势的重合率为多少。试验数据结果显示分类结果良好,识别准确度达到91.67%,达到要求。利用Matlab试验结果做出程序化交易系统趋势信号指标,并根据趋势指标进行程序化的交易信号优化与决策。

Whether it is technical or fundamental analysis,they all make the prediction to the market trend.The foundation of prediction is make a comparison to the previous trading conditions to the market,and it is believed that the same trend will happen again when the coincidence rate of the previous trading condition reaches to a certain extent.Based on a probabilistic neural network,the chosen data are the daily trading data of soya bean meal from Dalian Commodity Exchange as the research object.The basic trading trends in this paper are defined as five days trading data and quantify them as the PNN inputs.Via the PNN,nine main trading trends are classified.The tested data indicate that the identification result is good and the accuracy of trading trend identification reaches 91.67% and satisfies with experiment.The programmed trading trend index(TI)is obtained through the Matlab simulation result.The programmed trading system can optimize the trading signals and make a better decision based on the TI.