Basic methods of change-point detection of financial fluctuations

Financial market time series are usually approximated by random walks, however, we can easily find significant deviation from a simple random walk by analyzing high frequency market data. It is important to detect change points of potential statistical properties automatically from given time series. We apply Fisher's exact test for detection of trends in time series and show that the method works well for various types of temporal fluctuations. We show an example of application of this method for a foreign exchange market time series.