Application And Prediction Of Growth Enterprises Market (Gem) Index Based On Support Vector Machine

The companies listed on GEM, as Establishe d businesses, have some Commonness, for example they are established for a relatively shorter time, firm size are small, performance are not outstanding, etc. But most of companies listed on GEM are rapid growth companies with great potential for growth. G EM has a relatively lower IPO standard than other stock-market in China, the risk of GEM is relatively larger, it is a strictly regulated stock market, but it is also a science and technology incubator and cradle-type capital markets for high-growth businesses. The purpose of the establishment of the GEM is to serve the national strategy of independent innovation, GEM Index representing the general direction of China’s economic transformation. GEM-listed companies with an entrepreneurial, innovative, high-growth characteristics highly fit with the industrial restructuring and upgrading strategy. Thus, research on the GEM index can help understand the extent of the capital market acceptance of innovative start-ups, and the development of new industries, and i nvestment reference guide for investors. The use of artificial intelligence methods for predicting the index has been extensively studied, but forecast for the GEM index has not been studied, in particular involving use of support vector machine to predict GEM index. Forecast and analysis of GEM index for investment is very important. With the continuous development and growth of the GEM stock market, forecasts and discuss of the GEM index has important practical significance. Support vector machine model, in predicting GEM index, also has important guiding significance for investment and research.In this paper, I use single-step prediction method, using the before data to predict the next date of the closing index. First of all, I modeled PCA-PSO-LIBSVM model, and after data normalization, used principal component analysis to reduce the dimension of multidimensional data, and then applied the PSO to optimize parameters, and support vector machines gave regression forecast results. The entire data is divided into three parts: a declining market, volatility market and rising market. Respectively for each part I used PCA-PSO-LIBSVM model and Artificial Dynamic Neural Network model to predict and analyze, and depended on the research above I developed two tradin g strategy.Through predictive analysis and comparison I found that in the three of the aforementioned market PCA-PSO-LIBSVM support vector machine always obtain a more satisfactory prediction result than the Artificial Dynamic Neural Network model did. Then, using PCA-PSO-LIBSVM support vector machine model to predict the GEM Index, I build two T+0 trading strategies, including “the GEM ETF and the margin trading strategy” and “GEM index stock index futures T+0 trading strategy”, the two strategies has made good gains, reflecting the practicality prediction results.