def sort_bbn_nodes():
from bbn import BeliefNetwork
from dst import settings as dst_settings
from copy import copy, deepcopy
BBN = BeliefNetwork.from_bif(dst_settings.BBN_BIF)
blank_node = {
'cpt': {},
'parents': [],
'given': [],
'variables': [],
'tier': None
}
node_dict = {}
sorted_dict = {
'0': {}
}
for node_idx in BBN.probabilities:
node = BBN.probabilities[node_idx]
if not node_idx in node_dict.keys():
node_dict[node_idx] = deepcopy(blank_node)
node_dict[node_idx]['variables'] = BBN.variables[node_idx]
if node['given']:
node_givens = [(x, BBN.variables[x]) for x in node['given']]
else:
node_givens = []
node_dict[node_idx]['cpt'] = build_cpt_dict(node['cpt'], node_givens, BBN.variables[node_idx])
if node['given']:
node_dict[node_idx]['given'] = node['given']
for given_idx in node_dict[node_idx]['given']:
if not given_idx in node_dict.keys():
node_dict[given_idx] = deepcopy(blank_node)
node_dict[given_idx]['parents'].append(node_idx)
unsorted_nodes = node_dict.keys()
while len(unsorted_nodes) > 0:
unsorted_list = [x for x in unsorted_nodes]
for node_idx in unsorted_nodes:
if len(node_dict[node_idx]['parents']) == 0:
node_dict[node_idx]['tier'] = 0
unsorted_list.remove(node_idx)
sorted_dict['0'][node_idx] = node_dict[node_idx]
else:
for parent_idx in node_dict[node_idx]['parents']:
if not node_dict[node_idx]['tier']:
if not node_dict[parent_idx]['tier'] == None:
node_tier = node_dict[parent_idx]['tier']+1
node_dict[node_idx]['tier'] = node_tier
unsorted_list.remove(node_idx)
if not str(node_tier) in sorted_dict.keys():
sorted_dict[str(node_tier)] = {}
sorted_dict[str(node_tier)][node_idx] = node_dict[node_idx]
unsorted_nodes = unsorted_list
return sorted_dict
def test_from_bif(self):
bn = BeliefNetwork.from_bif(self.bif)
self.assertEquals(
sorted(list(bn.variables.keys())),
['Cancer', 'Dyspnoea', 'Pollution', 'Smoker', 'Xray'])
self.assertEquals(
sorted(list(bn.probabilities.keys())),
['Cancer', 'Dyspnoea', 'Pollution', 'Smoker', 'Xray'])
def update_bbn_bif(obj, post):
from bbn import BeliefNetwork
from dst import settings as dst_settings
BBN = BeliefNetwork.from_bif(dst_settings.BBN_BIF)
for field_key in post.keys():
if '-' in field_key:
node_key, condition_code = field_key.split('-')
if node_key in BBN.probabilities.keys():
field_val = post[field_key]
tuple_key = ()
if len(condition_code) > 0:
for idx, given_key in enumerate(BBN.probabilities[node_key]['given']):
tuple_key += (BBN.variables[given_key][int(condition_code[idx])],)
# assumes all variables are binary!!!
tuple_key_0 = tuple_key + (BBN.variables[node_key][0],)
tuple_key_1 = tuple_key + (BBN.variables[node_key][1],)
BBN.probabilities[node_key]['cpt'][tuple_key_0] = round(float(field_val), 2)
BBN.probabilities[node_key]['cpt'][tuple_key_1] = round(1-float(field_val), 2)
BeliefNetwork.to_bif(BBN, dst_settings.BBN_BIF)
obj.bif = get_bbn_nodes()
obj.save()
def systemcheck():
bn = BeliefNetwork.from_bif(settings.BBN_BIF)
# Valid BBN?
valid = bn.is_valid
if not valid[0]:
raise SystemCheckError("BBN from {} is not valid:\n {}".format(
settings.BBN_BIF, valid[1]))
# special case, contains a hidden terminal node for pit score
pitnode = '__overall_pit_restorability'
if pitnode not in bn.hidden_nodes:
raise SystemCheckError("Overall pit node `{}` required in bbn".format(pitnode))
# the terminal nodes of the belief network match questions
terminalnode_names = []
for name, prob in bn.probabilities.items():
if not prob['given'] and not name.startswith('__'):
# it's a terminal node or "hidden" node without questions
terminalnode_names.append(name)
from survey.models import Question
question_names = [q.name for q in Question.objects.all()]
tnn = set(terminalnode_names)
qn = set(question_names)
diff = list(qn - tnn)
if len(diff) > 0:
raise SystemCheckError(
"Extraneous questions without terminal nodes:\n\t%s" % diff)
diff = list(tnn - qn)
if len(diff) > 0:
raise SystemCheckError(
"Terminal nodes missing questions:\n\t%s" % diff)
# Make sure choices are valid;
# first choice in bn.variables must have value of 1.0
# all others must have values bt 0 and 1
for question in Question.objects.all():
first = bn.variables[question.name][0]
for choice in question.choices:
if choice['value'] < 0.0 or choice['value'] > 1.0:
raise SystemCheckError("`{}` has invalid values in choices "
"(must be 0 to 1)".format(question.name))
if choice['choice'] == first and choice['value'] != 1.0:
raise SystemCheckError("`{}` has invaid values in choices; "
"`{}` must be 1.0".format(question.name, first))
# randomly alter the probability table
# (first the True cond the false at 1.0 - val)
first = self.state.variables[var][0]
last = self.state.variables[var][1]
key = tuple(clist + [first])
lastkey = tuple(clist + [last])
val = self.state.probabilities[var]['cpt'][key]
newval = val + random.choice([0.05, -0.05])
if newval >= 0.0 and newval <= 1.0:
valid_move = True
self.state.probabilities[var]['cpt'][key] = newval
self.state.probabilities[var]['cpt'][lastkey] = 1.0 - newval
if __name__ == "__main__":
bn = BeliefNetwork.from_bif(INPUT_BIF)
training_sites = {}
sites_glob = os.path.abspath(
os.path.join(os.path.dirname(__file__),
'optimization_files', 'training_sites', "*.txt"))
user_datas = {}
for training_site in glob.glob(sites_glob):
socio_economic = site = landscape = None
user_data = {}
with open(training_site, 'r') as fh:
for line in fh.readlines():
key, val = line.split(',')
val = float(val)
try:
cond = bn.variables[key][0] # assume first
def test_to_bif(self):
bn = BeliefNetwork.from_bif(self.bif)
bn.to_bif(self.outbif)
bn2 = BeliefNetwork.from_bif(self.outbif)
self.assertEquals(bn2, bn)
from django.contrib.gis.db import models
from django.contrib.auth.models import User
from django.conf import settings
from jsonfield import JSONField
from django.core.validators import MaxValueValidator, MinValueValidator
from bbn import BeliefNetwork
BBN = BeliefNetwork.from_bif(settings.BBN_BIF)
class BaseModel(models.Model):
notes = models.TextField(default='', blank=True)
user = models.ForeignKey(User)
date_created = models.DateTimeField(auto_now_add=True)
date_modified = models.DateTimeField(auto_now=True)
class Meta:
abstract = True
def __str__(self):
return "{} `{}` by {}".format(self.__class__.__name__,
self.name,
self.user.username)
class BifSettings(BaseModel):
bif = models.TextField(default='', blank=True)
def __str__(self):
return"{} by {}".format(self.__class__.__name__,
self.user.username)
