Markov Chain Transition Diagrams in Python

Simple Markov Chain visualization module in Python. Only requires matplotlib and numpy to work.

Description

The current version works with 2 to 9 states - it is possible to plot more, but graph becomes cluttery. (2 to 4 states in the original version 2021-12-14.)

Getting Started

Dependencies

• matplotlib
• numpy

Installation

Copy the files src/node.py and src/markovchain.py in your script directory. Then

from markovchain import MarkovChain

2-state Markov chain demo

P = np.array([[0.8, 0.2], [0.1, 0.9]]) # Transition matrix
mc = MarkovChain(P, ['1', '2'])
mc.draw("../img/markov-chain-two-states.png")

Contribution by joel-n

4-state Markov chain demo

P = np.array([
    [0.65, 0.2, 0.15, 0.0], 
    [0.3, 0.3, 0.3, 0.15],
    [0.1, 0.0, 0.7, 0.15],
    [0.4, 0.2, 0.3, 0.1]
])
mc = MarkovChain(P, [r'$Z^1$', r'$Z^2$', r'$Z^3$', r'$Z^4$'])
mc.draw("../img/markov-chain-four-states.png")

Plotting more than 4 states with automatic coordinate generation

N = 7
P = np.random.random((N, N))
for n in range(N):
	P[n,:] = P[n,:]/np.sum(P[n,:])
P = np.around(P, 3)

mc = MarkovChain(P, [f'{i+1}' for i in range(N)])
mc.draw("../img/markov-chain-N-states.png")

Plotting HMM Markov chains with initial probabilities and emission means

N = 5
P = np.random.random((N, N))
for n in range(N):
	P[n,:] = P[n,:]/np.sum(P[n,:])
P = np.around(P, 3)

init_probs = np.random.random(N)
init_probs = np.around(init_probs / np.sum(init_probs), 3)
        
mean = np.random.random((N, 3))
        
mc = MarkovChain(P, [f'{i+1}' for i in range(N)], init_probs, mean)
mc.draw("../img/markov-chain-N-states-hmm.png")

Original author

Naysan Saran

Link to my blog.

License

This project is licensed under the GPL V3 licence.