def testNonAlignedArray(self):
'''Test the segmenter using an unaligned array'''
data = np.array(map(np.random.rand,(20,23,15)))
segment_parts = (3,2)
train_data, valid_data = timeSeriesSegmenter(data, segment_parts, validation_split=.5)
assert len(valid_data) > 0, 'Validation Data shouldn\'t be empty'
assert len(train_data) > 0, 'Training data shouldn\'t be empty'
def testOverlap(self):
'''Test the Time Series Segmenter overlap'''
data = np.random.rand(5,20)
segment_parts = (3,2)
train_data, valid_data = timeSeriesSegmenter(data,
segment_parts, allowable_overlap=2, validation_split=0)
assert len(train_data) == 30, 'Training data should be 100% of 30'
assert len(valid_data) == 0, 'Validation data should be 0% of 30'
def testNoSegmentsError(self):
'''Error case: No segment parts'''
try:
data = [[1,2]]
segment_parts = []
train, valid = timeSeriesSegmenter(data, segment_parts)
assert False, 'Error was not raised for an empty segment'
except AssertionError:
# This is good
pass
def testBasic(self):
'''Test the Time Series Segmenter basically'''
data = np.random.rand(5,20)
segment_parts = (3,2)
train_data, valid_data = timeSeriesSegmenter(data, segment_parts)
assert len(train_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(valid_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(train_data[0][0]) == 3, 'The first segment array should be 3'
assert len(train_data[0][1]) == 2, 'The second segment array should be 2'
def testMultiPartSegment(self):
'''Use a 4-part segment to segment the data'''
data = np.random.rand(5,27)
segment_parts = (3,2,3,1)
train_data, valid_data = timeSeriesSegmenter(data, segment_parts)
assert len(train_data) + len(valid_data) == 15, 'Total segments doesn\'t match'
assert len(train_data[0]) == 4, 'Each data tuple should have length equal to segment_parts'
assert len(train_data[0][0]) == 3, 'The first segment array should be 3'
assert len(train_data[0][1]) == 2, 'The second segment array should be 2'
assert len(train_data[0][2]) == 3, 'The third segment array should be 3'
assert len(train_data[0][3]) == 1, 'The fourth segment array should be 1'
def testSegmentOverlap(self):
'''Test the Time Series Segmenter segment overlap'''
data = np.random.rand(5,20)
segment_parts = (3,3)
train_data, valid_data = timeSeriesSegmenter(data,
segment_parts, allowable_overlap=0, validation_split=0, segment_overlap=(1,))
assert len(train_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(train_data[0][0]) == 3, 'The first segment array should be 3'
assert len(train_data[0][1]) == 3, 'The second segment array should be 3'
train_data = np.array(train_data)
assert (train_data[:,0,2] == train_data[:,1,0]).all(), "Segment overlap did not occur"
def testMultiPartSegmentWithOverlap(self):
'''Use a 4-part segment with overlaps to segment the data'''
data = np.random.rand(5,31)
segment_parts = (3,2,3,1)
allowable_overlap = 1
segment_overlap = (2,-1,1)
train_data, valid_data = timeSeriesSegmenter(data, segment_parts)
assert len(train_data) + len(valid_data) == 20, 'Total segments doesn\'t add up'
assert len(train_data[0]) == 4, 'Each data tuple should have length equal to segment_parts'
assert len(train_data[0][0]) == 3, 'The first segment array should be 3'
assert len(train_data[0][1]) == 2, 'The second segment array should be 2'
assert len(train_data[0][2]) == 3, 'The third segment array should be 3'
assert len(train_data[0][3]) == 1, 'The fourth segment array should be 1'
def testLargeSegmentation(self):
'''Test the Time Series Segmenter with a large dataset (overlap, segment overlap, multidim, and unaligned)'''
data = [np.random.rand(4,np.random.randint(100,2000)) for unused in range(200)]
segment_parts = (30,5)
segment_overlap = (1,)
allowable_overlap = 1
train_data, valid_data = timeSeriesSegmenter(data, segment_parts, allowable_overlap=allowable_overlap, segment_overlap=segment_overlap)
assert len(train_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(valid_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(train_data[0][0]) == 4, 'The first dimension of a segment should be the number of data types'
assert len(train_data[0][0][0]) == 30, 'The first segment array should be 30'
assert len(train_data[0][1]) == 4, 'The first dimension of a segment should be the number of data types'
assert len(train_data[0][1][0]) == 5, 'The second segment array should be 5'
assert len(train_data) > len(valid_data), 'Training Data should be much larger than validation data'
def testMultidimensionData(self):
'''Test the Time Series Segmenter using multi-dimension data'''
data = np.random.rand(5,6,20)
segment_parts = (3,2)
train_data, valid_data = timeSeriesSegmenter(data, segment_parts)
assert len(train_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(valid_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(train_data[0][0]) == 6, 'The first segment array should be of high dimension 6'
assert len(train_data[0][0][0]) == 3, 'The first segment array should be of low dimension 3'
assert len(valid_data[0][0]) == 6, 'The first segment array should be of high dimension 6'
assert len(valid_data[0][0][0]) == 3, 'The first segment array should be of low dimension 3'
assert len(train_data[0][1]) == 6, 'The second segment array should be of high dimension6'
assert len(train_data[0][1][0]) == 2, 'The second segment array should be of low dimension 2'
assert len(valid_data[0][1]) == 6, 'The second segment array should be of high dimension 6'
assert len(valid_data[0][1][0]) == 2, 'The second segment array should be of low dimension 2'
def testNonAlignedMultiDimArray(self):
'''Test the segmenter using a multidimensional unaligned array'''
doubleRand = lambda n: np.random.rand(2, n)
data = np.array(map(doubleRand,(20,23,15,25)))
segment_parts = (3,2)
train_data, valid_data = timeSeriesSegmenter(data, segment_parts, validation_split=.5)
assert len(valid_data) > 0, 'Validation Data shouldn\'t be empty'
assert len(train_data) > 0, 'Training data shouldn\'t be empty'
assert len(train_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(valid_data[0]) == 2, 'Each data tuple should have length equal to segment_parts'
assert len(train_data[0][0]) == 2, 'The first segment array should be of high dimension 2'
assert len(train_data[0][0][0]) == 3, 'The first segment array should be of low dimension 3'
assert len(valid_data[0][0]) == 2, 'The first segment array should be of high dimension 2'
assert len(valid_data[0][0][0]) == 3, 'The first segment array should be of low dimension 3'
assert len(train_data[0][1]) == 2, 'The second segment array should be of high dimension 2'
assert len(train_data[0][1][0]) == 2, 'The second segment array should be of low dimension 2'
assert len(valid_data[0][1]) == 2, 'The second segment array should be of high dimension 2'
assert len(valid_data[0][1][0]) == 2, 'The second segment array should be of low dimension 2'
