Research On Intelligent Modeling For Quantitative Stock Selection And Parameter Optimization

The quantitative stock selection model is an important part of the quantitative investment strategy and machine learning techniques are important research methods of the quantitative stock selection model.The improvement of the prediction accuracy of the model can help investors to maximize their profit.Since the stock market is a large nonlinear system,data preprocessing and parameters optimization algorithm will highly affect the model prediction performance.Therefore,traditional forecasting methods have been unable to study the comprehensive trend of stock price.In order to explore new method of quantitative stock selection,intelligent algorithm is applied,and a combination prediction model based on Principal Component Analysis(PCA),Artificial Bee Colony(ABC)algorithm,and Extreme Learning Machine(ELM)is proposed.1)Taking Shanghai and Shenzhen 300 shares as the research object,as all kinds of factors in the stock market have the dual characteristics of time and space,they are normalized according to their inherent economic logic.The normalization processing,the simple trend processing and the depth trend processing are carried out respectively.Then three groups of data are introduced into ELM and the prediction performance is compared with SVM.The results show that,after the data has been processed by the depth trend method,the learning of the positive sample is more sufficient,and the prediction performance of ELM to the rising stock is better and more practical.2)PCA algorithm is used to reduce the dimension of three groups of data,and then ELM is established.The experimental results show that PCA improves the prediction accuracy of the ELM model,and the superior accuracy of the depth trending data(53.61%)is exhibited.3)The ABC algorithm is combined with PCA-ELM and used to select best parameters combination for ELM.The optimized model is developed to predict the depth trending data,and the prediction performance are compared with PCA-GA-ELM and PCA-DE-ELM.The experimental results show that the ABC algorithm is better for the parameter optimization of ELM.In addition,Naive Bayes and Xgboost are used to predict the same data for further comparison.The results show that the prediction performance of PCA-ABC-ELM model is better and the prediction accuracy reached 59.67%.4)In order to further analyze the benefits of the model,the prediction results of PCA-ABC-ELM are transformed into investment signals,then the short-term and medium-term simulation investment experiments are carried out on the Shanghai and Shenzhen 300 index component stocks.The experimental results show that the 4-day cumulative income is 8189.11 yuan and the annual rate of return is 66.38%;the 21-day cumulative income is 50835.63 yuan and the annual rate of return is 80.45%.Both of them are significantly superior to the performance of the Shanghai and Shenzhen 300 index in the same period,which shows the feasibility and effectiveness of PCA-ABC-ELM model in the quantitative stock selection of Shanghai and Shenzhen 300 shares.