def finalTest(individual):
func = toolbox.lambdify(expr=individual)
correct = 0.0
trainingResult = [ kernelCalc(func,t) for t in instancesTraining ]
d0 = [ r for (r, l) in zip(trainingResult, labelsTraining) if l == 0 ]
d1 = [ r for (r, l) in zip(trainingResult, labelsTraining) if l == 1 ]
from findOptimum import linearSearch, bruteForce
v = linearSearch(d0, d1)
print "Training: ", v
value = bruteForce(d0, d1)
print "Optimum Value = ", value
#TODO: if the result is the other way around, I should check it here also
result = []
test = [ kernelCalc(func,t) for t in instancesTest ]
for t in test:
if t < value:
result.append(0)
else:
result.append(1)
correct = sum( not (a ^ b) for (a, b) in zip(result, labelsTest) )
total = len(instancesTest)
print "Accuracy = ", correct / total
print "Accuracy (scikit) = ", metrics.accuracy_score(labelsTest, result)
print "F1-score (micro) = ", metrics.f1_score(labelsTest, result, average='micro')
print "F1-score (macro) = ", metrics.f1_score(labelsTest, result, average='macro')
print "F1-score (weighted) = ", metrics.f1_score(labelsTest, result, average='weighted')
print metrics.classification_report(labelsTest, result)
fitness = correct/total
return fitness, #len(individual)
def evaluate(individual, metric_):
#TODO: check how can I separate the two datasets...
func = toolbox.lambdify(expr=individual)
result = [ kernelCalc(func,t) for t in instancesTraining ]
#resultLabel = [ (r, l) in zip(result, labelsTraining) ]
#print "result=", result
d0 = [ r for (r, l) in zip(result, labelsTraining) if l == 0 ]
d1 = [ r for (r, l) in zip(result, labelsTraining) if l == 1 ]
#print "D0 = ", len(d0)
#print "D1 = ", len(d1)
#print individual
from findOptimum import linearSearch
fitness = linearSearch(d0, d1)
fitness -= (len(individual)**2)/100.0
#print "fit=", fitness
#if metric_ == "f1":
# fitness = metrics.f1_score(labelsTraining, result)
#elif metric_ == "acc":
# total = len(instancesTraining)
# fitness = correct / total
return fitness, #len(individual)
