def customizeFeatureLabel(strDataPath,
lsFileNames,
lsDataAxisName,
dSamplingFreq,
dcLabel,
nResponsePerData,
nSegmentPerRespsonse,
dVibrationDuration,
dIntervalDuration,
dRestDuration,
lsValidSegmentID,
dValidSegmentDuration,
lsAxis2Inspect = ['x0', 'y0', 'z0'] ):
"""
Extract customized features and labels from give data set
Parameters:
----
strPath:
folder path of data
lsFileNames:
names of files to use
dcLabel:
diction for mapping btw file name and label
nResponsePerData:
number of responses in single data
nSegmentPerRespsonse:
number of segments in single response
dVibrationDuration:
duration of each vibration segment
dIntervalDuration:
static duration btw segments
dRestDuration:
static duration btw responses
lsValidSegmentID:
list of segments to extract features
dValidSegmentDuration:
duration of segment to extract features
lsAxis2Inspect:
axis to extract features, default is ['x0', 'y0', 'z0']
Returns:
----
dfFeatures:
a data frame of features
lsLabels:
a list of corresponding labels
"""
if (len(lsValidSegmentID) == 0 or
len(lsValidSegmentID) > nSegmentPerRespsonse or
dValidSegmentDuration <= 0.0 or
dValidSegmentDuration > dVibrationDuration):
raise ValueError()
# load data
lsData = md.loadDataEx(strDataPath, lsFileNames, lsDataAxisName)
# extract features & labels
lsDataFeatures = []
lsLabels = []
for strDataName, dfData in zip(lsFileNames, lsData):
nLabel = dcLabel[strDataName[:2] ]
lsAxisFeatures = []
# extract features for each axis
for strCol in lsAxis2Inspect:
dcAxisFeatures = customizeAxisFeature(dfData,
strDataName,
strCol,
dSamplingFreq,
nResponsePerData,
nSegmentPerRespsonse,
dVibrationDuration,
dIntervalDuration,
dRestDuration,
lsValidSegmentID,
dValidSegmentDuration)
if (len(dcAxisFeatures.keys() ) != nResponsePerData ):
raise RuntimeError("the responses extracted "
"does not equal to predefined")
dfAxisFeatures = pd.DataFrame(dcAxisFeatures).T
lsAxisFeatures.append(dfAxisFeatures)
# combine X, Y, Z horizontally
dfDataFeatures = pd.concat(lsAxisFeatures, axis=1,
ignore_index=False)
# save for furthur combination
lsDataFeatures.append(dfDataFeatures)
lsLabels.extend([nLabel]*dfDataFeatures.shape[0])
# concatenate feature frames of all data together (vertically)
dfFeatureLabel = pd.concat(lsDataFeatures, axis=0, ignore_index=False)
dfFeatureLabel[LABEL] = pd.Series(lsLabels, index=dfFeatureLabel.index)
return dfFeatureLabel
def extractFeatureLabel(strDataPath,
lsFileNames,
lsDataAxisName,
dSamplingFreq,
dcLabel,
nResponsePerData,
nSegmentPerRespsonse,
dVibrationDuration,
dIntervalDuration,
dRestDuration,
lsAxis2Inspect = ['x0', 'y0', 'z0']):
"""
Extract features and labels from give data set
Parameters:
----
strPath:
folder containing the data
lsFileNames:
list of data file names
dcLabel:
the diction of label mapping
nResponsePerData:
number of responses in a data
Returns:
----
dfFeatureLabel:
a data frame of features & corresponding labels
"""
# load data
lsData = md.loadDataEx(strDataPath, lsFileNames, lsDataAxisName)
# extract features & labels
lsDataFeatures = []
lsLabels = []
for strDataName, dfData in zip(lsFileNames, lsData):
nLabel = dcLabel[strDataName[:2] ]
lsAxisFeatures = []
# extract features for each axis
for strCol in lsAxis2Inspect:
dcAxisFeatures = extractAxisFeature(dfData,
strDataName,
strCol,
dSamplingFreq,
nResponsePerData,
nSegmentPerRespsonse,
dVibrationDuration,
dIntervalDuration,
dRestDuration)
if (len(dcAxisFeatures.keys() ) != nResponsePerData ):
raise RuntimeError("the responses extracted does not "
"equal to predefined")
dfAxisFeatures = pd.DataFrame(dcAxisFeatures).T
lsAxisFeatures.append(dfAxisFeatures)
# combine X, Y, Z horizontally, note the index is used!
dfDataFeatures = pd.concat(lsAxisFeatures, axis=1,
ignore_index=False)
# save for furthur combination
lsDataFeatures.append(dfDataFeatures)
lsLabels.extend([nLabel]*dfDataFeatures.shape[0])
# concatenate feature frames of all data together (vertically)
dfFeatureLabel = pd.concat(lsDataFeatures, axis=0, ignore_index=False)
dfFeatureLabel[LABEL] = pd.Series(lsLabels, index=dfFeatureLabel.index)
return dfFeatureLabel