# Test classifier
testSet = DataUtils.read_dataset(directory +"Set-"+ setNo +"-validate.csv")
predictionsRaw = classifier.predict(testSet.features)
# Outbox labels
predictions = []
predictionsExpl = []
for row in range(len(predictionsRaw)):
# Convert float to int
predictions.append(int(predictionsRaw[row]))
# Splice int into 3 values
order = predictions[row]/100
sat = (predictions[row]/10)%10
gran = predictions[row]%10
predictionsExpl.append([order, sat, gran])
# Write results to csv file
results = DataUtils.ResultSet(testSet.id, predictionsExpl, testSet.labels)
DataUtils.write_resultset(results, directory +"Result-"+ setNo +".csv")
# Metrics of classifier
import Metrics
matrix_complete = Metrics.ConfusionMatrix(predictions, inboxLabels(testSet.labels))
# Readable printing
#Metrics.printConfusionMatrix("COMPLETE", matrix_complete)
# Simple printing
Metrics.printConfusionMatrixCompact(matrix_complete)
print ""
print DataUtils.removeMultipleRows(range(len(names)),
missingColumns), "\t",
# Cropped training and validation data
cropFeatures = DataUtils.removeMultipleColumns(features,
missingColumns)
cropValidation = DataUtils.removeMultipleColumns(
testSet.features, missingColumns)
# Select classifier
classifier = svm.SVC()
# Train classifier
classifier = classifier.fit(cropFeatures, labels)
# Test classifier
predictionsRaw = classifier.predict(cropValidation)
# Outbox labels
predictions = []
for row in range(len(predictionsRaw)):
# Convert float to int
predictions.append(int(predictionsRaw[row]))
# Metrics of classifier
matrix_complete = Metrics.ConfusionMatrix(predictions, testLabels)
# Simple printing
Metrics.printConfusionMatrixCompact(matrix_complete)
print ""
#print cropFeatures[0]
import Metrics
import pandas as pd
confusion_matrix = Metrics.ConfusionMatrix(14)
Metrics.ConfusionMatrix.count_predicted(1, 1)
v = torch.max(outputs.data, 1)[1]
unique, counts = np.unique(labels_test, return_counts=True)
dict(zip(unique, counts))
for i in range(len(list(model.parameters()))):
print(list(model.parameters())[i].size())
#Nuumber of parameters in out model -------------------------------------------
p = 1
s = 0
for i in range(len(parameters)):
p = 1
for j in range(len(parameters[i].size())):
p = parameters[i].size(j) * p
s = s + p
if ((epoch - (epoch % 100)) // 100) % 2 == 0:
old_ratio[epoch % 100] = loss.data[0]
else:
new_ratio[epoch % 100] = loss.data[0]
if (epoch % 200 == 0) and (epoch > 200):
str_p = ""
str_a = ""
for col in range(dim):
p_single = int(p[col])
a_single = int(a[col])
str_p += str(p_single) + " "
str_a += str(a_single) + " "
predict_single[col].append(p_single)
actual_single[col].append(a_single)
str_p = str_p[0:-1]
str_a = str_a[0:-1]
predict_complete.append(str_p)
actual_complete.append(str_a)
import Metrics
matrix_complete = Metrics.ConfusionMatrix(predict_complete, actual_complete)
matrix_single = []
for col in range(dim):
matrix_single.append(
Metrics.ConfusionMatrix(predict_single[col], actual_single[col]))
# Hardcoded for better formatting
#Metrics.printConfusionMatrix("COMPLETE", matrix_complete)
#Metrics.printConfusionMatrix("EXPLORATION ORDER", matrix_single[0])
#Metrics.printConfusionMatrix("SATURATION STRATEGY", matrix_single[1])
#Metrics.printConfusionMatrix("SATURATION GRANULARITY", matrix_single[2])
# Simple printing
Metrics.printConfusionMatrixCompact(matrix_complete)
for col in range(dim):
c = ["Cat"] d = ["Dog"] r = ["Rabbit"] import Metrics predict = 7 * c + 8 * d + 12 * r actual = 5 * c + 2 * d + 3 * c + 3 * d + 2 * r + d + 11 * r confMatrix = Metrics.ConfusionMatrix(predict, actual) # Basic stats: print "Matrix:" print confMatrix.matrix print "Groups:", confMatrix.index print "Number of groups:", confMatrix.size print "Number of tests:", confMatrix.quantity # Advanced stats: confMatrix.createSubMatrices() print "" print "Recall (macro) :", confMatrix.recallMacro() print "Precision (macro):", confMatrix.precisionMacro() print "Accuracy (macro) :", confMatrix.accuracyMacro() print "Accuracy (micro) :", confMatrix.accuracyMicro()
