SVM Method Based On Kernel Parameter Optimization And Its Application In Stock Market

The stock market plays an important role in the economic development of the country and plays an important role in economic development.The research on the prediction of stock prices is of great significance both to individual and corporate investors and relevant government departments in formulating relevant economic policies.With the development of the field of statistical machine learning,various intelligent algorithms are also emerging.From the point of view of the characteristics of stock prices,it has great uncertainty in the short term,but in the long-term trend is consistent with the statistical law.Therefore,in the case of finite samples,predicting stock prices through machine learning algorithms is an important direction in the stock prediction research.Among them,the support vector machine(SVM)method has a strong advantage.It does not require much sample size,can solve non-linear problems,and has strong promotion ability.Therefore,in this paper,based on the support vector machine method,a prediction model for kernel parameters optimization of support vector machines is established based on feature extraction,and good prediction results are obtained.The specific research content is as follows:1.For the index data of the Shanghai Composite Index,the PCA method was used to analyze the principal components of the original data,and 46 technical indicators were used as input features to perform correlation analysis.High-relevancy features were removed after the correlation analysis to reduce the dimension,and a total of 754 days of indicator data was normalized.After processing,labeling was carried out,and then five main components with a cumulative variance contribution rate of more than 85%and 18 main components of nearly 99%were used as the input feature data of the support vector machine by principal component analysis.2.In the process of support vector machine training,the key parameters of the SVM were adjusted using genetic algorithm,particle swarm optimization algorithm and artificial fish swarm algorithm respectively,and a more accurate support vector machine model was obtained.Among them,the particle swarm algorithm has the fastest convergence speed,and the artificial fish swarm algorithm performs better.3.Based on the advantages of fast convergence speed and good convergence effect of artificial fish swarm algorithm,this paper establishes a support vector machine prediction model based on accelerated artificial fish swarm algorithm,and uses the speed update rules of artificial swarm algorithm to artificial fish swarm algorithm.The convergence speed is optimized,and the empirical research results show that among the above four algorithms,the SVM prediction model based on the accelerated artificial fish school algorithm performs best,and the cross-check accuracy rate is 79.02 when the PCA is used.%,81.54%when PCA is not used,and it takes 179.01s.