Event Effect Based Financial Time Series Prediction

The computer is drilling down into all areas of society, network information is the indispensable source of information, but its non-structured, sources dispersal features makes deep network information processing technology develops slowly, at the same time, people have used structured storage of quantitative data, compared to network information which have a structured, accurate, stable characteristics of the source. At present, these two fields are still independent , because of the accuracy of quantitative data, making it playing a decisive role in time series forecasting,but purely based on structured data time series prediction model is difficult to predict the mutagenic effects of events from the Internet. As an exploration, this paper attempts to fuse these two types of independent information.The article first study the ERBF kernels SVM classifier in financial time series prediction. In order to classify the trend of stock linearly, the main difficulty is how to select features and kernel functions. At last, by choosing 4-day stock technical analysis data feature and ERBF function, build a better classifier than similar prediction algorithm.On this basis, we study the effect of special events how to apply to the stock market forecast. As a result of the financial markets is very complex, structured time series prediction is difficult to react to emergencies. This article use the method based on that the event how much affect the stock, classify the event, and calculate the confidence of the event, according to the confidence, combined with the structured time series prediction, establish the time series prediction based on the effects of the information, can deal with the effects of information which the structured time series prediction cannot do.Finally, this article use a lot of experiments to test the effect of the above methods, the results show that, based on ERBF kernel SVM classifier predict the trends of structural financial time series, the accuracy rate is 58.9%, which is better than other similar methods, special events based on financial time series forecasting accuracy rate is as high as 75.3%, achieve the desired results.