Research Of Financial Volatility Based On Sequence Alignment Method

With intensified fluctuation of the financial market and its wide spread in theglobal world, it is of great importance to analyze and forecast the volatility offinancial market with scientific method. Financial metrology has made abundantachievements in the research of financial fluctuation. However the economic systemis a complicated nonlinear system, the rule of the financial volatility can’t be graspedin all aspects just with the financial metrology method. New methods are needed tostudy the fluctuation problem from different angles. As a supplement of the financialmetrology research method, symbolic time series analysis method and sequencealignment method just can do this.Through introduction of sequence alignment method in bioinformatics andnonparametric symbolic time series analysis, a new method of forecasting financialvolatility is put forward in this paper combined with the existing K-Nearest Neighbormethod. Sequence comparison is made between the alignment goal sequence andsample sequence with the time series data after symbolization. K historicalsub-sequences whose matching score are higher than the threshold value are backedout using the dynamic programming algorithm and viewed as K nearest neighbor inK-NN method. Their weights are calculated respectively in order to get the predictingoutcomes. Empirical analysis of price fluctuation sequence is made with the highfrequency data from Shanghai Stock Exchange Composite Index and Shenzhen StockExchange Component Index. On the basis of a single price fluctuation variable,forecast is extended to two variables at the same time including fluctuation of thetransaction price and transaction time interval, fluctuation of the transaction price andtrading volume. Two dimensional time series is turned into one dimensional symbolicseries with appropriate symbolic method. The feasibility and validity of this methodare proved through empirical analysis which is made with ultra-high frequency date ofVanke stock. This method can capture nonlinear characteristics of the time sequenceand reduce the noise sensitivity. There is no need to find out which model the datageneration process is in accordance with and to make stable hypothesis of the data.This method can predict not only the specific fluctuation value but also the fluctuationinterval and it can be used widely. The first chapter of this paper states the background and research status of thefinancial market fluctuation at home and abroad and puts forward innovative points inthe text. The second chapter summarizes two important basic theory of symbolic timeseries method and sequence alignment methods. The third chapter puts forward theforecasting method and detailed steps based on the sequence alignment method.Empirical analysis is made with high frequency data whose sampling interval is20minutes from Shanghai Stock Exchange Composite Index and Shenzhen StockExchange Component Index. Then the fluctuation time series and fluctuationsymbolic sequence are forecasted respectively. The fourth chapter extends forecastfrom the single variable to two variables and makes empirical analysis with ultra-highfrequency data from Vanke stock in March2010. The fifth chapter summarizes thefull work in the text. It points out that the sequence alignment method still has greatapplication and development space in the research of financial market and alsoindicates the direction for the next research and improvement.