search

Enderlogic / MMHC-Python

An implementation of MMHC in python
13 stars 4 forks source link

bug #5

Open Joker222565 opened 7 months ago

Joker222565 commented 7 months ago

代码

Joker222565 commented 7 months ago

import random import urllib.request from os import path

import pandas as pd from graphviz import Digraph from rpy2.robjects import pandas2ri from rpy2.robjects.packages import importr

from lib.evaluation import f1 from lib.mmhc import mmhc

pandas2ri.activate() base, bnlearn = importr('base'), importr('bnlearn')

load network

network = 'ecoli70' network_path = 'Input/' + network + '.rds' if not path.isfile(network_path): url = 'https://www.bnlearn.com/bnrepository/' + network + '/' + network + '.rds' urllib.request.urlretrieve(url, network_path) dag_true = base.readRDS(network_path)

generate data

datasize = 1000 filename = 'Input/' + network + '_' + str(datasize) + '.csv' if path.isfile(filename): data = pd.read_csv(filename, dtype='float64') # change dtype = 'float64'/'category' if data is continuous/categorical else: data = bnlearn.rbn(dag_true, datasize) data = pd.DataFrame(data) data = data[random.sample(list(data.columns), data.shape[1])] data.to_csv(filename, index=False)

learn bayesian network from data

dag_learned = mmhc(data)

plot the learned graph

dot = Digraph() for node in bnlearn.nodes(dag_learned): dot.node(node) for parent in bnlearn.parents(daglearned, node): dot.edge(node, parent) dot.render('output/' + network + '' + str(datasize) + '.gv', view = False)

evaluate the learned graph

print('f1 score is ' + str(f1(dag_true, dag_learned))) print('shd score is ' + str(bnlearn.shd(bnlearn.cpdag(dag_true), dag_learned)[0]))

binxinyangcd commented 1 month ago

The same situation. May I ask do YOU fix the bug?