This note is purely in English for my final exam, please be calm down if you are in trouble.
Lets use Markov Chain to solve a very simple problem - 'Weather Problem', we have the following assumptions:
- If it is raining today, then the probability of raining tomorrow is 70%, and the probability of not raining tomorrow is 30% ;
- If it is not raining today, then the probability of raining tomorrow is 40%, and the probability of not raining tomorrow is 60% .
Therefore, we have the following finite state machine:
According to the state machine, we can get the one-step transition matrix as follows:
![](https://i-blog.csdnimg.cn/blog_migrate/684ad6c7750de472faf4e3fb4a218420.png)
Now, lets use the stationary probability of Markov Chain, for computing the statistical probability of raining or not raining for someday.
- At first, we make the staring distribution in Day 0 is
![](https://i-blog.csdnimg.cn/blog_migrate/e52422cb054fe279726950653d5b1f76.png)
- Then, the distribution in Day 1 is
![](https://i-blog.csdnimg.cn/blog_migrate/5fe827c14b64cd6816af56b37e4d0d63.png)
- Then, the distribution in Day 2 is
![](https://i-blog.csdnimg.cn/blog_migrate/02dd21d894267fab4c80c7401017ae26.png)
- Finally, for given transition matrix and the staring distribution, the stationary distribution can be computed as
![](https://i-blog.csdnimg.cn/blog_migrate/a23677a88d2a2a7dd3f5667cfd0d2158.png)
Actually, we have already got our goal, the final stationary distribution can describe the statistical probability.
In conclusion, just need 3 steps for analysis of practical example using Markov Chain.
- Draw the State Machine through the existing conditions;
- Compute the one-step transition matrix;
- Compute the stationary distribution according to the existing conditions.