def runA1andA2(K, dynamic_datasets_path):
dataSets = readDatasets.read_dataset(True, True, False,
dynamic_datasets_path)
trainSet = dataSets[0]
testSet = dataSets[1]
maxWarpingWindowPercentage = 0.11
makeListOfAllNeighbors = False
plotPatterns = True
task2A1.findKnearestNeighbors(K, maxWarpingWindowPercentage, plotPatterns,
makeListOfAllNeighbors, trainSet, testSet)
# This doesn't take any dataSets as parameters.. it gets them on its own
task2A2.runLCSS(K, dynamic_datasets_path)
def runClassification(K, dynamic_datasets_path):
print('Start KNN Classification..')
dataSets = readDatasets.read_dataset(True, False, True,
dynamic_datasets_path)
trainSet = dataSets[0]
testSet = dataSets[1]
makeListsOfNeighborsForAllTests = True
plotPatterns = False # We just want the KNN, not the html-maps.
maxWarpingWindowPercentage = 0.33 # For testSet_a2, we need a bigger window to get the right patternIDs.
### Run KNN for test_a2
neighborsTestsLists = findKnearestNeighbors(
K, maxWarpingWindowPercentage, plotPatterns,
makeListsOfNeighborsForAllTests, trainSet, testSet)
testData = getVotes(neighborsTestsLists)
write_predictions_to_csv(testData)
### Run cross-validation.
crossValidation(trainSet, K, maxWarpingWindowPercentage, num_folds=10)
def crossValidation(trainSet, K, maxWarpingWindowPercentage, num_folds):
accuracies = []
subset_size = int(len(trainSet) / num_folds)
for i in range(num_folds):
testing_this_round = trainSet[i * subset_size:][:subset_size]
training1 = trainSet[0:][:subset_size * i]
training2 = trainSet[(i + 1) * subset_size:][:subset_size *
(10 - i - 1)]
training_this_round = training1 + training2 # np.concatenate((training1, training2), axis=0)
# print training2['journeyPatternId']
makeListsOfNeighborsForAllTests = True
plotPatterns = False # We just want the KNN, not the html-maps.
neighborsTestsLists = findKnearestNeighbors(
K, maxWarpingWindowPercentage, plotPatterns,
makeListsOfNeighborsForAllTests, training2, testing_this_round)
testData = getVotes(neighborsTestsLists)
print(testData)
correct = 0
i = 0
for row in trainSet['journeyPatternId']:
if i == 100:
break
try:
print("Predicted: ", testData[i][1], ' - ', "Actual: ", row)
if testData[i][1] is row:
correct += 1
except IndexError:
break
i += 1
print('Correct predictions: ', correct)
accuracy = float(correct) / 100
print('Accuacy: ', accuracy)
return accuracy