改进隐马尔可夫模型在日降雨发生模拟中的应用  

Application of improved Hidden Markov Model in simulation daily rainfall occurrence

作　　者：张蕊[1] 夏乐天[1] 

机构地区：[1]河海大学理学院,江苏南京211100

出　　处：《人民长江》2012年第20期17-20,共4页Yangtze River

摘　　要：一般的隐马尔可夫模型(HMM)无法有效地在观察和隐藏过程中纳入季节性变量。将一般的隐半马尔可夫模型(HSMM)加以改进,允许马氏链状态的任意停留时间分布。应用R函数,建立了包含季节性的停留时间分布的HMM模型,模拟了河北省各城市日降雨的发生。模拟结果表明,特别是在隐藏的马氏链中,该模型可以有效地在观察过程和隐藏过程中纳入季节性变量;普通HMM模型的状态预测、解码和模型检验的公式和方法也同样适用于该模型。The seasonal variable can not be incorporated into the general Hidden Markov Model(HMM) during hidden and observed process effectively.The general Hidden semi-Markov Models(HSMMs) is improved,which allows arbitrary dwell-time distributions of Markov state.The R function is applied and the HMM that contains seasonal dwell-time distribution is established to simulate the daily rainfall occurrence in cities of Hebei Province.The results show that the seasonal variable can be incorporated into the hidden and observed process effectively,especially in the hidden Markov chain;the formulas and methods for state prediction,decoding and check of general HMM can also be used in the presented model as well.

关 键 词：隐马尔可夫模型 隐半马尔可夫模型 停留时间分布 最大似然估计 日降雨发生 

分 类 号：O212.2[理学—概率论与数理统计]