GA优化的SVM在量化择时中的应用

Application of SVM Optimized by Genetic Algorithm in Quantization Timing Selection

针对量化投资过程中因交易信号判断不准确而导致的择时难问题,利用具有优良非线性可分能力的支持向量机建立基于历史价量信息(开盘价、收盘价、最高价、最低价、成交量和短长期移动平均指数)的量化择时模型.在策略模型的具体应用中,为了确定LIBSVM Tool Box中的"-c"和"-g"参数,本文首先通过遗传算法对其寻优,然后利用MATLAB软件实现了对个股(浦发银行)自2012年1月4日至2016年6月22日的策略回测,最后以沪深300指数为基准从年化收益率、相关绩效指标和最大回撤等角度对回测结果进行了分析,得出GA-SVM可被有效运用到量化择时中去的结论.

For quantitative investment caused by inaccurate trading signal judgment during the process of the timing of difficult problems,the excellent non-linear separable ability is used to support vector machine（SVM）based on historical price quantity information（opening price,closing price,the highest and the lowest price,volume and short long term moving average）model of quantitative timing.In the specific application of strategy model,in order to determine LIBSVM ToolBox in the ＆quot;c＆quot; and ＆quot;g＆quot; parameter,this paper optimize them through the genetic algorithm,then uses MATLAB software to achieve the（Shanghai pudong development bank）for individual stocks from January 4,2012 to 2012 on January 22,the strategy of back,finally the csi 300 index as the benchmark from the annualized yield,sharpe ratio,the angle of information ratio and maximum retracement back to the measurement results are analyzed.It is concluded that the GA-SVM can more accurately judge the conclusion of trading signals.