中文摘要：

将小波变换与符号时间序列分析相结合，引入工程领域的D-Markov模型，提出了一种用于金融波动变化模式识别和异常检测的方法。波动序列经过离散小波变换，产生小波系数序列，将小波系数序列符号化产生符号时间序列，建立符号时间序列的D-Markov模型，并求状态转移概率矩阵，计算各状态转移概率矩阵的状态概率向量与标准状态转移概率矩阵的状态概率向量之间的欧拉距离，从而得到异常度。基于得到的异常度识别金融波动变化模式，检测异常波动的发生。以上证综指的5分钟序列为样本实证分析，对该方法的可行性和有效性进行了验证。

英文摘要：

Through the introduction of D -Markov model in the field of engineering , combined with the wavelet transform and symbolic time series analysis , a method of financial volatility pattern recognition and anomaly detection was put forward . Wavelet coefficient sequence was generated from the discrete wavelet transform of volatility series .Then, the D-Markov model of symbolic time series , which was the wavelet coefficient sequence after symbolization , was build to get the state transition prob-ability matrix.Subsequently , anomaly measure as the standard Euclidean distance between state probability vectors of states tran -sition probability matrixes and state probability vector under standard volatility series was computed .Financial volatility pattern was recognized and abnormal volatility was detected based on the anomaly measure .With high frequency data whose sampling in-terval was 5 minutes from Shanghai Stock Exchange Composite Index , the feasibility and validity of this method was proven through empirical analysis .