class InputReaderTest(unittest.TestCase):
def setUp( self ):
''' Construct our InputReader object, pass it the test csv file '''
self.mInputReader = InputReader( testFile )
def test_read( self ):
''' Test the csv read functionality against known values '''
self.mInputReader.readFile()
mRawData = self.mInputReader.getRawData()
self.assertEqual( mRawData, [['InputReader', 'Test', 'CSV'],
['1', '44', '-4.3'],
['234', '-45', '0.45']] )
def test_pathSet( self ):
''' Test set file path and FileNotFoundError exception'''
# Set a bogus file name
self.mInputReader.setFilePath( '../../res/NonExist.csv' )
self.assertRaises( FileNotFoundError, self.mInputReader.readFile() )
class FeatureExtractor( metaclass=ABCMeta ):
'''
Abstract base class for extracting and generating features from input
resource file. Implementation classes must implement getFeatures()
and getTrainingData() appropriately for the given input source.
'''
def __init__( self, mInputReader , filterPath='../res/FeatureFilter' ):
'''
Constructor - arguments passed from main
@param mInputReader: InputReader object for setting raw data
'''
# Feature dump and filter path
self.outCSVPath = '../../tmp/featureDump.csv'
self.filterCSVPath = filterPath
# Get raw data from the passed InputReader
mInputReader.readFile()
self.rawData = mInputReader.getRawData()
# Initialize feature set and training data from raw data
self.features = self.rawData[0]
self.trainingData = np.array( self.rawData[1:] )
# Construct the InputReader used for feature filtering
self.filterReader = InputReader( self.filterCSVPath )
# Initialize number of samples removed
self.nRmvSamples = 0
def setOutCSVPath( self , fPath ):
'''@param fPath: relative location and name of feature dump CSV'''
self.outCSVPath = fPath
def setFilterPath( self , fPath ):
'''@param fPath: relative location and name of feature filter CSV'''
self.filterCSVPath = fPath
def getFeatures( self ):
return self.features
def getTrainingData( self ):
return self.trainingData
def setTrainingData( self, data ):
assert( isinstance( data, np.ndarray ) )
self.trainingData = data
def getSampleCnt( self ):
return len( self.trainingData )
def getRmvSampleCnt( self ):
return self.nRmvSamples
def listIdx( self, feature ):
'''
Return the list index of a given feature
@param feature: training feature
@return index: index of passed feature
'''
return self.features.index( feature )
def applyFeatureFilter( self ):
'''
Reads the filter resource file and accordingly removes the feature
from each sample.
'''
# Read out the resource content
self.filterReader.readFile()
# Stash the results to a local list
mFilterList = self.filterReader.getRawData()[0]
# Use our list index method to find appropriate column in feature
# list to remove
for feature in mFilterList:
try:
idx = self.listIdx( feature )
del self.features[idx]
self.trainingData = np.delete( self.trainingData, idx, 1 )
except ValueError:
print( 'Unable to remove feature %s!' % feature )
def writeFeaturesToCSV( self ):
'''
Dump the transformed data out to CSV for external eval
Note: This shouldn't be called w/o extracting features from a
derived class first.
'''
mDumpFile = open( self.outCSVPath, 'w', newline='' )
mCSVWriter = csv.writer( mDumpFile, delimiter=',' )
# First write the features to the first row of the dump file
mCSVWriter.writerow( self.features )
# Then, dump all training data writing by row/record
mCSVWriter.writerows( self.trainingData )
# Release file i/o
mDumpFile.close()
@abstractmethod
def extractFeatures( self ):
''' This method is to be implemented by subclasses'''
pass
def __del__( self ):
'''No Destructor implementation'''
pass
