The Prediction Of Stock’s Crucial Turning Point Based On Polynomial-fitting And Support Vector Machine

In Western countries, quantitative investment after more than three decades of development, the application of the quantitative investment in the investment market is very extensive. In the Chinese stock market, the concept of quantitative investment has just been introduced. Because of the special environment for the development of the Chinese market, because of its own superiority, quantitative investment has a very broad prospects.This article is based on the theory of quantitative investment, and different from most of the traditional method to predict the stock’s complete future price curve, turn the prediction of the stock price curve into a problem whether the stock’s price is in the turning state. It is expected that the machine learning algorithm can be used to train a model which can determine whether the stock price is in the upward or downward direction at the current time. When the stock price is in the crucial upward turning point, you can buy stock to achieve profitability, when the stock in the crucial downward turning point, you can sell stocks stop loss, I hope the model can provide some decision-making advice to investors in the stock.How to get the high efficiency and reliability samples in the historical data by the computer before the training of the support vector machine classification model is the key point. In order to solve disadvantages of the traditional identification method in threshold selection and the effect of the turning point. In this paper, I propose a new method to identify the crucial turning point in discrete data key based on polynomial-fitting, The contrast experiments show that the new method has better results.In the support vector machine in the process of training, this article regards the MACD, KDJ and turnover rate of stock price which reflected the stock’s trend as feature vector. We apply the method of the key points of the discrete data set based on polynomial fitting to the analysis of these techniques, and generate the corresponding feature vectors. Then I select the RBF kernel function to learning, use GA and cross validation method, to optimize two optional parameters of radial basis kernel function, and obtain the best classification model. Use the above steps several times Combined Training Technical indicators to elect a best combination of features and classification up two key inflection point based on the combination of features and training generated classification model down a key turning point. And then use the above steps several times to train the data set, and choice the best feature group and the best model.Eventually, after a lot of experiment compared to the traditional prediction, the prediction of stock’s crucial turning point based on SVM and polynomial-fitting has higher accuracy. And it can provide some investment advices to investors.