NEEQ Financial Time Series Model Based On Support Vector Machine And Application

With the continuous development of the multi-level capital market in Chi-na, stock market is becoming an important boost of economic development, carrying the financing needs of enterprises and the investment needs of people. The bull market since 2014 and the turbulence of Chinese stock market from the second half of 2015 to now made us have to re-examine the Chinese stock market and the entire capital market. At the same time, the NEEQ market which was so-called "the Chinese NASDAQ" has becoming an important basic component of the multi-level capital market in China. Now it is a market fac-ing to institutional investor due to its relatively special status, investor access threshold and trading mode, and the characteristics of its market data are somewhat different from the Shanghai and Shenzhen market. Because of its short development time, the market indexes have only came out for nearly one year. For all aspects of the market research, it is not very complete.The SVM method which is based on statistical learning and structural risk minimization principle is one of the more effective methods of intelligent prediction.lt has unique advantages in solving the problem of data mining with small sample, poor information, nonlinear and so on. The two main applied forms of SVM-SVC and SVR have been applied in financial market as powerful research tools and they provide a way for us to research the NEEQ market. In this paper, the empirical research is based on the SVM method and the main research object is the NEEQ market making index, mainly from two aspects.The first one is to choose some indicators which are highly relating to the morrow opening price sequence. We select appropriate kernel function ac-cording to data feature and use the appropriate method to choose appropriate parameters. Then we train a v-ε-SVR model to predict the morrow opening price sequence of the NEEQ market making index.Another one is to select some representative technical indicators and set sliding window to mine the short-term market trend reversal points identified by these technical indicators. We use the days’closing prices which have been smoothed to define the true reversal points. Then we mark and collect all the "suspected reversal point" and divide them into two parts-the training set and the testing set. After this we choose the kernel function and select the best parameter combination. Finally we train the C-SVC model and predict the reversal points in the testing set to provide references to investors to determine the short-term market changes.This paper makes an empirical study on the NEEQ market making index by using the SVM model in different forms and has achieved quite ideal perfor-mance. It has some reference value for us to attend the NEEQ market-making transaction and even to make some further studies.