Research On The Analysis And Predictive Algorithm Of Financial Time Series Based On Support Vector Machine

Modern financial sector has become the important driving force for socio-economic development of a country and an important part of national competitiveness. It’s more directly related to the stability of the financial markets, efficiency and security, that understand and grasp the essential rules of the financial markets. In view of the important role of the stock index return series and volatility series in portfolio and risk aversion, scientifically predicting financial market volatility characteristics to master the volatility of financial markets law and its structure to avoid the risk prevention and management and monitoring is of great significance.The main stock index as a barometer of the world economy, its revenue analysis and volatility forecasting is the precondition and foundation of many financial model studies. Therefore, the research of the index returns and volatility as subjects has financial significance.Time series in scientific researching and engineering are analyzed by time series modeling. Time series modeling is to create a mathematical model to intuitively reflect the relationship between the input and output data. The establishment of the process for such a mathematical model that is the process of prediction using of the data obtained: using the already known data as input, the resulting output is the prediction of future data. Therefore, how to find or construct efficient and accurate prediction model is the hot research topic.This study generally includes the following sections:Based on the phase space reconstruction theory of support vector regression analysis to predict the financial markets. The focus of phase space reconstruction is to select an appropriate delay time, thereby reducing the necessary embedding dimension.In this paper," the false nearest neighbor point method "is used,and build the error function for all variables to determine the appropriate embedding dimension combination. Kernel functions play important role in support vector regression. Support vector regression based on phase space reconstruction theory has a relatively good earnings forecast performance.LSSVR-CARRX model algorithm.The LSSVR algorithms replace the standard linear error term by using a quadratic error term, and solve linear quadratic programming problem with equality constraints instead of the computation of quadratic programming problems with inequality constraints. Transformed equivalent to the dual problemequations.The LSSVR advantage is to significantly improve the accuracy and training speed. In this paper,a nonlinear CARRX model based the the LSSVR is constructed.The model use the LSSVR powerful nonlinear mapping ability, to establish nonlinear relationship between the non-linear function of output and input. The model fitting and forecasting of volatility.Stock market prediction algorithm based on neural networks and continuous relevance vector machine neural network Hopfield network. Network stability and convergence is a prerequisite for the Hopfield network, energy function is focus around this prerequisite. Continuous Hopfield network has its own limitations, resulting in multiple pseudo corresponding affect the calculation accuracy. The core concepts of relevance vector machine (RVM) is the introduction of a chance to explain the noise. Research and testing is to prove:the relevance vector data advantage. In order to improve the accuracy of the relevance vector machine algorithm, this paper presents an improved algorithm:The HRVM algorithm, the feedback network problem into the objective function and constraints, and then construct an energy function, so that this optimization problem boils down to satisfy the constraints, the minimum of the objective function. The HRVM improves algorithm with high stability, and can be applied in trend forecasting system.Financial time series prediction is based on support vector regression algorithm. The support vector calculation method set field of machine learning standard as one of the organic integration of a variety of techniques and methods used to solve some challenging application problems, a wider range of applications, but also able to achieve the best results can not be achieved by other methods, butis also facing some difficulties and problems,such as:the kernel function and parameter selection to provide theoretical guidance of the mountain, the need to find large-scale data fast training and testing methods.Based on this, this paper presents an adaptive learning Support Vector Machine regression algorithm.The algorithm based on Support Vector Machine algorithm to further improve and optimize the convex quadratic programming problem with a single linear equations instead of solving, improve the speed of the machine learning training. Experimental results show that:the algorithm is scientific and effective exchange rate time series data to complete short-term prediction In addition, the fund price forecast experimental results show that the accuracy of SAL-SVR is higher than SVR.In response to the nonlinear characteristics of the financial system, Chaos theory and Support Vector Machine theory are applied in the research of multivariate financial time series. The results show that these theories could effectively nonlinearly analyze and forecast financial market volatility.