算法：HMM模型+维特比算法详解

一、HMM模型+维特比算法实例

1、问题描述

假设连续观察3天的海藻湿度为(Dry,Damp,Soggy),求这三天最可能的天气情况。

2、已知信息

①天气只有三类(Sunny,Cloudy,Rainy)，海藻湿度有四类{Dry,Dryish, Damp,Soggy }，而且海藻湿度和天气有一定的关系。

②隐藏的状态：Sunny, Cloudy, Rainy;

③观察状态序列：{Dry, Damp, Soggy}

④初始状态序列：

Sunny	Cloudy	Rainy
0.63	0.17	0.20

⑤状态转移矩阵：

	Sunny	Cloudy	Rainy
Sunny	0.5	0.375	0.125
Cloudy	0.25	0.125	0.625
Rainy	0.25	0.375	0.375

⑥发射矩阵：

	Dry	Dryish	Damp	Soggy
Sunny	0.6	0.2	0.15	0.05
Cloudy	0.25	0.25	0.25	0.25
Rainy	0.05	0.10	0.35	0.5

3、分析

　　由一阶HMM可知，Day2的天气仅取决于Day1；Day3的天气又只取决于Day2的天气。

4、计算过程

（1）Day1由于是初始状态，我们分别求

P(Day1-Sunny)=0.63*0.6;

P(Day1-Cloudy)=0.17*0.25;

P(Day1-Rain)=0.20*0.05;

Choose max{ P(Day1-Sunny) , P(Day1-Cloudy),P(Day1-Rainy)}, 得到P(Day1-Sunny)最大，得出第1天Sunny的概率最大。

（2）Day2的天气又取决于Day1的天气状况，同时也受Day2观察的海藻情况影响。

P(Day2-Sunny)= max{ P(Day1-Sunny)*0.5, P(Day1-Cloudy)*0.25,  P(Day1-Rainy)*0.25} *0.15;

P(Day2-Cloudy)= max{ P(Day1-Sunny)*0.375,  P(Day1-Cloudy)*0.125, P(Day1-Rainy)*0.625} *0.25;

P(Day2-Rainy)= max{ P(Day1-Sunny)*0.125,  P(Day1-Cloudy)*0.625 , P(Day1-Rainy)*0.375} *0.35;

Choosemax{ P(Day2-Sunny) , P(Day2-Cloudy), P(Day2-Rainy)},得到P(Day2-Cloudy)最大，得出第2天Cloudy的概率最大。

故{Sunny,Cloudy}是前两天最大可能的天气序列。

（3）Day3的天气又取决于Day2的天气状况，同时也受Day3观察的海藻情况影响。

　　P(Day3-Sunny)= max{ P(Day2-Sunny)*0.5, P(Day2-Cloudy)*0.25,  P(Day2-Rainy)*0.25} *0.05;

　　P(Day3-Cloudy)= max{ P(Day2-Sunny)*0.375,  P(Day2-Cloudy)*0.125, P(Day2-Rainy)*0.625} *0.25;

　　P(Day3-Rainy)= max{ P(Day2-Sunny)*0.125,  P(Day2-Cloudy)*0.625, P(Day2-Rainy)*0.375} *0. 05;

　　Choosemax{ P(Day3-Sunny) , P(Day3-Cloudy), P(Day3-Rainy)},得到P(Day3-Rainy)最大，得出第3天Rainy的概率最大。故{Sunny,Cloudy,Rainy}是这三天最可能的天气序列。

5.Python代码

流程图：

#-*- coding:utf-8 -*-
__author__ = 'Administrator'

init_vec={"sunny":0.63,"cloudy":0.17,"rainy":0.20}
trans_mat={"sunny":{"sunny":0.5,"cloudy":0.375,"rainy":0.125},
           "cloudy":{"sunny":0.25,"cloudy":0.125,"rainy":0.625},
           "rainy":{"sunny":0.25,"cloudy":0.375,"rainy":0.375}}
emit_mat={"sunny":{"dry":0.6,"dryish":0.2,"damp":0.15,"soggy":0.05},
          "cloudy":{"dry":0.25,"dryish":0.25,"damp":0.25,"soggy":0.25},
          "rainy":{"dry":0.05,"dryish":0.10,"damp":0.35,"soggy":0.50}}
observes=["dry","damp","soggy"]
states=["sunny","cloudy","rainy"]

#列表中包含字典，就相当于二阶列表使用
tab=[{}] #只有一行
path=[{}]
for t in range(len(observes)):
    #print(t)
    if t==0:
        temp=[]
        for state in states:
            prob=init_vec[state]*emit_mat[state].get(observes[t])
            tab[0][state]=prob
            temp.append((prob,state))
        best_prob,best_state=max(temp,key=lambda x:x[0])
        path[0][best_state]=best_prob
    else:
        tab.append({})
        path.append({})
        temp=[]
        for state1 in states:
            item=[]
            for state2 in states:
                prob=tab[t-1][state2]*trans_mat[state2].get(state1)*emit_mat[state1].get(observes[t])
                item.append((prob,state2))
            best_prob,best_state=max(item,key=lambda x:x[0])
            tab[t][state1]=best_prob
            temp.append((best_prob,state1))
        best_prob,best_state=max(temp,key=lambda x:x[0])
        path[t][best_state]=best_prob
print(tab)
print(path)

输出结果：

[{'rainy': 0.010000000000000002, 'cloudy': 0.0425, 'sunny': 0.378}, {'rainy': 0.0165375, 'cloudy': 0.0354375, 'sunny': 0.02835}, {'rainy': 0.01107421875, 'cloudy': 0.0026578125, 'sunny': 0.0007087500000000001}]

[{'sunny': 0.378}, {'cloudy': 0.0354375}, {'rainy': 0.01107421875}]