Research On Chinese Stock Market Price Forecast Based On Online Learning Support Vector Machine

The traditional support vector machine algorithm processes data in a batch manner,and it repeatedly classifies all data.When faced with the huge amount of data frequently updated in the stock market today,it may lead to higher time space resource requirements.The core idea of online learning is to break through the traditional batch-based learning method.Through the continuous updating of new data,the model is trained and continuously tuned to reflect the data changes quickly and accurately,and improve the prediction accuracy.By reading the literature,we find that the online classification of such support vector machines is partially applied in biological,pharmaceutical,engineering and other projects,but it has not been widely spread in the financial market.Therefore,this paper studies the combination of online learning and support vector machine algorithms to predict the stock market.In this paper,the related literatures are studied first.The theory and implementation of online learning are reviewed.The methods of applying support vector machines to financial markets are reviewed and summarized.Compared to the common support vector machine(SVM),the least squares support vector machine(LSSVM)converts the inequality constraint into an equality constraint,thereby reducing the Lagrangian multiplier alpha.The solution complexity has faster solution efficiency,but it also loses the sparseness of solution.Therefore,in order to solve this problem,this paper chooses the online learning method to improve the least squares support vector machine and constructs the model through mathematical derivation.Using the new algorithm model,the LSSVM is tested and compared with the past stock price information of China Merchants Bank.The optimal kernel parameters and kernel function,the radial basis kernel function,are obtained through training,and the improved model prediction effect is better than LSSVM.According to the selected optimal kernel function,the optimal parameters selection and demonstration of 9 stocks in the three industries of large,medium and small discs are carried out through the online learning LSSVM algorithm.Finally,the paper summarizes the empirical results and the full text content:(1)The new model can better grasp the future trend of stocks;(2)The forecasting effect of the new model has a certain delay;(3)The fluctuation of the predicted value is larger than the actual value.(4)There is a difference in the prediction between stocks of different magnitudes.At the same time,some suggestions are put forward for the future application and expansion of the model:(1)expanding the type and scope of sample selection;(2)using different feature vectors,data processing,and kernel parameter optimization to verify the model adaptability;Build quantitative methods and investment strategies based on the new model.