Researching Probability Density Prediction Method Through Quantile Regression Neural Network And Kernel Density Estimation

Probability density prediction method is a method based on statistical and probability theory, which has a scientific theory in statistics and probability theory. Through the complete graph of the probability density, we not only can get a more accurate prediction of the point value to provide more accurate quantitative basis for management decisions; but also can get a complete continuous probability density curve and its probability density function of the prediction object. Thus, through the analysis of the probability density curve of random variables, It can provide more accurate and useful information for management decisions.In this paper, the neural network quantile regression method, not only can overcome the shortcomings of least-squares regression, but also can combine with Neural Network’s adaptive and strong nonlinear advantage. By setting different sub sites, we predict the random variable’s different quintile point corresponding quintile, which can be interpreted as a more detailed portrait of the variables affected the explanatory variables and the relationship between them in the different sub-sites. By using neural network quantile regression forecasting method, we find objects sites on different points corresponding quantile, and then, we have come to a complete forecast object continuous probability density curve with the form of the kernel density estimation and optimal window width selection methods and the thinking of the kernel density estimation method. By adding different factors, and choosing a different form of kernel density estimation and optimal window width selection methods, we can be obtained a prediction target that probability density curve and obtain the prediction error is also very different when we compare with different considerations and different forms of kernel density estimate and optimal window width selection methods. The peper presented quantile regression neural network and kernel density estimation method to predict the probability density on the basis of an integrated neural network quantile regression and kernel density estimation method, and then, the probability of a specific grid electricity load forecasting corporate density empirical research and stock price probability density empirical studies using this method, The probability density of neural network based on quantile regression and kernel density estimation in-depth study and exploration,which enrich and develop better prediction theory methodology.