Asthma

In [1]:

import matplotlib.pyplot as plt

# import the computation tools of aGrUM
import pyAgrum as gum

# import the graphical display functions
import pyAgrum.lib.notebook as gnb


The model

In [2]:

# load the "asthma" Bayesian network

In [3]:

# display the Bayesian network
gnb.showBN(bn,nodeColor={n:0.9 for n in bn.names()},cmapNode=plt.get_cmap('Blues'))

In [4]:

# display the conditional probability table of asthme given pollution
gnb.showPotential(bn.cpt(bn.idFromName('asthma')),digits=4)

asthma
pollution
crisis
no_crisis
1
0.10000.9000
2
0.21130.7887
3
0.19720.8028
4
0.28240.7176
5
0.47460.5254
6
0.52300.4770
7
0.63410.3659
8
0.82570.1743
9
0.85370.1463
10
0.90910.0909

Some inference

In [5]:

# display the probability distribution of Variable "trafic"
gnb.showPosterior (bn, {}, "traffic" )

In [6]:

# display the probability distribution of Variable "pollution"
gnb.showPosterior ( bn, {}, "pollution")

In [7]:

# display the distribution somewhat differently
gum.getPosterior ( bn, target="pollution")

Out[7]:

pollution
1
2
3
4
5
6
7
8
9
10
0.00000.03550.42330.27880.14750.06290.01640.02190.00820.0055
In [8]:

# more interesting: display the posterior distribution of "asthme"
# given that we observed that time is 8:00 and weather is cloudy
gnb.showPosterior (bn, evs={'hour' : 8, 'weather' : 'cloudy'}, target="asthma" )

In [9]:

# show the complete model on morning (from 8 to 12)
gnb.showInference(bn,evs={'hour' : [0 if i< 8 or i> 12 else 1 for i in range(24)]})

In [10]:

# show the posterior distributions of all the variables given that
# we observed that heure=8 and meteo=nuageux.
# the tables in beige represent the observations
gnb.flow.row(gnb.getInference(bn,size="9",evs={'hour' : 8, 'weather' : 'cloudy'}),
gnb.getInference(bn,size="9",evs={'hour': 7, 'accident' : 'yes'}),
captions=["When time is 8:00 and weatcher is cloudy","When time is 7:00 and there hase been an accident"])


When time is 8:00 and weatcher is cloudy

When time is 7:00 and there hase been an accident