def find_best_feature_partition(examples,
compare_fun,
feature,
max_splits):
assert isinstance(examples, ml_exampleset)
assert type(feature) == IntType
assert type(max_splits) == IntType
best_relinfgain = -1
best_absinfgain = -1
best_part = None
part_gen = partition_generator_feature(examples, feature, max_splits)
while 1:
try:
part = part_gen.next()
apriori = part.get_class_entropy()
cost = part.get_entropy()
remainder = part.get_remainder_entropy()
# print "# A-priori, cost, remainder = (%2.6f, %2.6f, %2.6f)" %\
# (apriori,cost,remainder)
absinfgain = apriori-remainder
relinfgain = pylib_probabilities.rel_info_gain(apriori,
remainder,
cost)
#print relinfgain, absinfgain
if compare_fun((relinfgain, absinfgain),
(best_relinfgain,best_absinfgain)) > 0:
best_relinfgain = relinfgain
best_absinfgain = absinfgain
best_part = part
except StopIteration:
break
return (best_relinfgain, best_absinfgain, best_part)
def find_best_feature_partition(examples, compare_fun, feature, max_splits):
assert isinstance(examples, ml_exampleset)
assert type(feature) == IntType
assert type(max_splits) == IntType
best_relinfgain = -1
best_absinfgain = -1
best_part = None
part_gen = partition_generator_feature(examples, feature, max_splits)
while 1:
try:
part = part_gen.next()
apriori = part.get_class_entropy()
cost = part.get_entropy()
remainder = part.get_remainder_entropy()
# print "# A-priori, cost, remainder = (%2.6f, %2.6f, %2.6f)" %\
# (apriori,cost,remainder)
absinfgain = apriori - remainder
relinfgain = pylib_probabilities.rel_info_gain(
apriori, remainder, cost)
#print relinfgain, absinfgain
if compare_fun((relinfgain, absinfgain),
(best_relinfgain, best_absinfgain)) > 0:
best_relinfgain = relinfgain
best_absinfgain = absinfgain
best_part = part
except StopIteration:
break
return (best_relinfgain, best_absinfgain, best_part)
def plain_rel_inf_gain(self):
apriori_entropy = pylib_probabilities.compute_entropy_absdistrib([1]
* self.get_class_number())
real_entropy = self.get_class_entropy()
return pylib_probabilities.rel_info_gain(apriori_entropy,
real_entropy,
real_entropy)
def entropies(self):
"""
Return a tuple of various entropic measures.
"""
tp = tree_partition(self, self.sample)
apriori = tp.get_class_entropy()
cost = tp.get_entropy()
remainder = tp.get_remainder_entropy()
absinfgain = apriori-remainder
if cost == 0:
relinfgain = 0
else:
relinfgain = pylib_probabilities.rel_info_gain(apriori,
remainder,
cost)
return (apriori, remainder, absinfgain, relinfgain)
def plain_rel_inf_gain(self):
apriori_entropy = pylib_probabilities.compute_entropy_absdistrib(
[1] * self.get_class_number())
real_entropy = self.get_class_entropy()
return pylib_probabilities.rel_info_gain(apriori_entropy, real_entropy,
real_entropy)
