在python中为libpgm包使用json格式文件

Question

I'm a new programmer with python. I want to use bayesian network for classification. Therefore i'm using libpgm package in python.The first codes for using this package is:

from libpgm.nodedata importNodeData
from libpgm.graphskeleton importGraphSkeleton
from libpgm.discretebayesiannetwork importDiscreteBayesianNetwork
from libpgm.pgmlearner importPGMLearner

nd=NodeData()
skel=GraphSkeleton()
fpath="json.txt"
nd.load(fpath)
skel.load(fpath)

The problem is that i don't know anything about content of this json file and path. I have installed Ananconda on my windows. Is there anyone who can help me? Is there anyone who have such file for a small bayesian network such as asia? Thanks.

Answer 1

This GitHub repo seems to contain some usage examples and data.

The Code example can be found here . Try to make it work locally.

Answer 2

The Bayesian network has to be represented as a valid JSON. So if you have a sufficiently big network, check the JSON formatting (eg using JSONlint) before wasting your time in debugging. Else nd.load() will give errors.

If the network loads but you still get errors is CPD factorization (while using TableCPDFactorization()), check the data again. Eg there should be no space after commas. Even though it will be a valid JSON it will not work due to indexing into the Vdataentry in discretebayesiannetwork.py. Also check that parents are in correct order in the tables and probabilities are correct.