"""
from abraxas4.abraxasFrame import AbraxasFrame
xgbAbra = False
dtAbra = True
if xgbAbra:
xgbAbra = AbraxasFrame(numIrSensors=10,
numFrSensors=2,
windowWidth=100,
windowShift=10,
numCoeffs=5,
numFreqs=1,
enaStatFeats=True,
wavelet='haar',
waveletLvl1=False,
featNormMethod='stand',
trainFraction=0.66,
classSortTT=True,
randomSortTT=False,
lineThresholdAfterNorm=10,
enaRawFeats=False,
corrPeaks=2)
xgbAbra.selectSensorSubset(selectedSensors=[False, True, True],
sensorType='bno')
xgbAbra.addDataFiles(fileSourceName="igor.txt",
fileSourcePath="../",
startTime=3550,
stopTime=3800,
from scipy.spatial.distance import euclidean
from abraxasOne.helperFunctions import writeMatrixToCsvFile
from abraxasThree.classifierClass import AbraxasClassifier
from fastdtw import fastdtw
from abraxasOne.gaussFilter import gaussFilter
from tscAlgs.triang import triang
from tscAlgs.dtwImp02 import dtwImp02
from abraxas4.abraxasFrame import AbraxasFrame
b = AbraxasFrame(numIrSensors=10,
numFrSensors=2,
windowWidth=250,
windowShift=50,
numFreqs=0,
numCoeffs=0,
enaStatFeats=False,
featNormMethod='none',
trainFraction=2 / 3,
waveletLvl1=False,
randomSortTT=False,
classSortTT=True)
b.setWindowFunction(functionName='rect', alpha=0.25)
# b.plotWindowFunction()
b.selectSensorSubset(selectedSensors=[False, False, False], sensorType='bno')
b.selectSensorSubset(selectedSensors=[], sensorType='fr')
b.selectSensorSubset(selectedSensors=[0, 1, 2, 3, 5, 7, 9], sensorType='ir')
b.addDataFiles(fileSourceName="igor2.txt",
fileSourcePath="../",
"""
cb, 07.08.2018
- user identification (with non walking data) using decision tree alg
"""
from abraxas4.abraxasFrame import AbraxasFrame
dtAbra = AbraxasFrame(numIrSensors=10,
numFrSensors=2,
windowWidth=100,
windowShift=25,
numFreqs=1,
numCoeffs=0,
enaStatFeats=True,
featNormMethod='stand',
trainFraction=2 / 3,
waveletLvl1=False,
randomSortTT=False,
classSortTT=True,
corrPeaks=0,
enaRawFeats=False)
dtAbra.loadTeTrDump(dumpName='dtAbra.pkl')
from sklearn.tree import DecisionTreeClassifier
clf = DecisionTreeClassifier(criterion="entropy",
max_depth=20,
min_samples_leaf=1,
min_samples_split=3,
max_features=None,
"""
cb, 07.08.2018
- user identification (with non walking data) using xgboost
"""
from abraxas4.abraxasFrame import AbraxasFrame
xgbAbra = AbraxasFrame(numIrSensors=10,
numFrSensors=2,
windowWidth=100,
windowShift=25,
numCoeffs=5,
numFreqs=1,
enaStatFeats=True,
wavelet='haar',
waveletLvl1=False,
featNormMethod='stand',
trainFraction=0.66,
classSortTT=True,
randomSortTT=False,
lineThresholdAfterNorm=10,
enaRawFeats=False,
corrPeaks=2)
xgbAbra.loadTeTrDump(dumpName='xgbAbra.pkl')
#from xgboost import XGBClassifier
import xgboost
clf = xgboost.XGBClassifier(max_depth=3, learning_rate=0.4)
xgbAbra.trainClassifier(classifier=clf, supervised=True)
xgbAbra.testClassifier()
from abraxas4.abraxasFrame import AbraxasFrame
import numpy as np
oc = AbraxasFrame(numIrSensors=10,
numFrSensors=2,
windowWidth=100,
windowShift=10,
numFreqs=1,
numCoeffs=3,
enaStatFeats=True,
featNormMethod='stand',
trainFraction=2 / 3,
waveletLvl1=False,
randomSortTT=False,
classSortTT=True,
corrPeaks=0,
enaRawFeats=False,
statStages=3)
oc.setWindowFunction(functionName='tukey', alpha=0.1)
oc.selectSensorSubset(selectedSensors=[False, True, True], sensorType='bno')
# oc.selectSensorSubset(selectedSensors=[], sensorType='fr')
# oc.selectSensorSubset(selectedSensors=[0], sensorType='ir')
oc.addDataFiles(fileSourceName="chrisOut2.txt",
fileSourcePath="../",
startTime=1500,
stopTime=5000,
label=0)
oc.addDataFiles(fileSourceName="chrisOut2.txt",
def getRabeData(sensors, length, shift=256):
from abraxas4.abraxasFrame import AbraxasFrame
b = AbraxasFrame(numIrSensors=10, numFrSensors=2, windowWidth=length, windowShift=shift, numFreqs=0, numCoeffs=0,
enaStatFeats=False, featNormMethod='none', trainFraction=2/3, waveletLvl1=False,
randomSortTT=False, classSortTT=True)
b.setWindowFunction(functionName='rect', alpha=0.25)
# b.plotWindowFunction()
b.selectSensorSubset(selectedSensors=[False, False, False], sensorType='bno')
b.selectSensorSubset(selectedSensors=[], sensorType='fr')
b.selectSensorSubset(selectedSensors=sensors, sensorType='ir')
#b.selectSensorSubset(selectedSensors=[2], sensorType='ir')
b.addDataFiles(fileSourceName="igor2.txt", fileSourcePath="../", startTime=600, stopTime=6000, label=0)
b.addDataFiles(fileSourceName="ankita_pos2_lrRl.txt", fileSourcePath="../", startTime=150, stopTime=2500, label=1)
b.addDataFiles(fileSourceName="markus.txt", fileSourcePath="../", startTime=500, stopTime=3300, label=2)
dataSet = b.readDataSet(checkData=False, equalLength=True)
# dataSet := Array with shape dataSet[i][j, k], where i refers to the i-th file loaded, k indicates the sensor and
# j is the "time"-index.
wData, wLabels = b.windowSplitSourceDataTT(inputData=dataSet, inputLabels=np.array([0, 1, 2]))
wLabels = np.array(wLabels)
print("Number of windows, Igor: ", str(np.size(wLabels[wLabels == 0])))
print("Number of windows, Ankita: ", str(np.size(wLabels[wLabels == 1])))
print("Number of windows, Markus: ", str(np.size(wLabels[wLabels == 2])))
igor = []
ankita = []
markus = []
#for i in range(len(wLabels)):
# wData[i] = wData[i]*(1+0*np.random.random([250, 1]))
for i in range(len(wLabels)):
if wLabels[i]==0:
igor.append(wData[i]*1)
if wLabels[i]==1:
ankita.append(wData[i]*1)
if wLabels[i]==2:
markus.append(wData[i]*1)
return igor, ankita, markus
from abraxas4.abraxasFrame import AbraxasFrame
import numpy as np
oc = AbraxasFrame(numIrSensors=10,
numFrSensors=2,
windowWidth=2**8,
windowShift=2**5,
numFreqs=0,
numCoeffs=0,
enaStatFeats=True,
featNormMethod='stand',
trainFraction=2 / 3,
waveletLvl1=False,
randomSortTT=False,
classSortTT=True,
corrPeaks=0,
enaRawFeats=False,
statStages=4)
oc.setWindowFunction(functionName='tukey', alpha=0.9)
oc.selectSensorSubset(selectedSensors=[False, True, True], sensorType='bno')
# oc.selectSensorSubset(selectedSensors=[], sensorType='fr')
# oc.selectSensorSubset(selectedSensors=[0], sensorType='ir')
oc.addDataFiles(fileSourceName="igor.txt",
fileSourcePath="../",
startTime=100,
stopTime=1500,
label=0,
className="walking")
"""
cb, 07.08.2018
- user identification (with non walking data) using decision tree alg
"""
from abraxas4.abraxasFrame import AbraxasFrame
gNbAbra = AbraxasFrame(numIrSensors=10, numFrSensors=2, windowWidth=50, windowShift=50, numFreqs=1, numCoeffs=1,
enaStatFeats=True, featNormMethod='minmax', trainFraction=2/3, waveletLvl1=False, randomSortTT=False,
classSortTT=True, corrPeaks=0, enaRawFeats=False)
gNbAbra.setWindowFunction(functionName='tukey', alpha=0.1)
gNbAbra.selectSensorSubset(selectedSensors=[False, False, False], sensorType='bno')
gNbAbra.selectSensorSubset(selectedSensors=[0, 1, 2], sensorType='ir')
gNbAbra.selectSensorSubset(selectedSensors=[], sensorType='fr')
'''
gNbAbra.addDataFiles(fileSourceName="igor.txt", fileSourcePath="../", startTime=3550, stopTime=3800, label=0,
className="not walking")
gNbAbra.addDataFiles(fileSourceName="igor2.txt", fileSourcePath="../", startTime=300, stopTime=500, label=0)
gNbAbra.addDataFiles(fileSourceName="ankita.txt", fileSourcePath="../", startTime=0, stopTime=150, label=0)
gNbAbra.addDataFiles(fileSourceName="markusSchnell.txt", fileSourcePath="../", startTime=4100, stopTime=4300, label=0)
gNbAbra.addDataFiles(fileSourceName="stefan.txt", fileSourcePath="../", startTime=7600, stopTime=8600, label=0)
gNbAbra.addDataFiles(fileSourceName="stefan.txt", fileSourcePath="../", startTime=0, stopTime=300, label=0)
gNbAbra.addDataFiles(fileSourceName="ben.txt", fileSourcePath="../", startTime=0, stopTime=1000, label=0)
gNbAbra.addDataFiles(fileSourceName="ben.txt", fileSourcePath="../", startTime=7100, stopTime=8000, label=0)
gNbAbra.addDataFiles(fileSourceName="chris1.txt", fileSourcePath="../", startTime=5200, stopTime=6000, label=0)
gNbAbra.addDataFiles(fileSourceName="novcc.txt", fileSourcePath="../", startTime=0, stopTime=10000, label=0)
gNbAbra.addDataFiles(fileSourceName="nowalk.txt", fileSourcePath="../", startTime=0, stopTime=10000, label=0)
from abraxas4.abraxasFrame import AbraxasFrame
import numpy as np
import random
import xgboost
from sklearn.model_selection import StratifiedKFold
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import train_test_split
#abra = AbraxasFrame(numIrSensors=10, numFrSensors=2, windowWidth=100, windowShift=10, numCoeffs=5, numFreqs=1,
# enaStatFeats=True, wavelet='haar', waveletLvl1=False, featNormMethod='stand', trainFraction=0.66,
# classSortTT=True, randomSortTT=False, lineThresholdAfterNorm=10, enaRawFeats=False, corrPeaks=2)
abra = AbraxasFrame(numIrSensors=10, numFrSensors=2, windowWidth=100, windowShift=10, numFreqs=1, numCoeffs=5,
enaStatFeats=True, featNormMethod='stand', trainFraction=2/3, waveletLvl1=False,
randomSortTT=False, classSortTT=True, corrPeaks=1, enaRawFeats=False)
abra.setWindowFunction(functionName='tukey', alpha=0.3)
abra.selectSensorSubset(selectedSensors=[False, False, False], sensorType='bno')
abra.addDataFiles(fileSourceName="igor.txt", fileSourcePath="../", startTime=3550, stopTime=3800, label=0,
className="not walking")
abra.addDataFiles(fileSourceName="igor2.txt", fileSourcePath="../", startTime=300, stopTime=500, label=0)
abra.addDataFiles(fileSourceName="ankita.txt", fileSourcePath="../", startTime=0, stopTime=150, label=0)
abra.addDataFiles(fileSourceName="markusSchnell.txt", fileSourcePath="../", startTime=4100, stopTime=4300, label=0)
abra.addDataFiles(fileSourceName="stefan.txt", fileSourcePath="../", startTime=7600, stopTime=8600, label=0)
abra.addDataFiles(fileSourceName="stefan.txt", fileSourcePath="../", startTime=0, stopTime=300, label=0)
abra.addDataFiles(fileSourceName="ben.txt", fileSourcePath="../", startTime=0, stopTime=1000, label=0)
abra.addDataFiles(fileSourceName="ben.txt", fileSourcePath="../", startTime=7100, stopTime=8000, label=0)
abra.addDataFiles(fileSourceName="chris1.txt", fileSourcePath="../", startTime=5200, stopTime=6000, label=0)
abra.addDataFiles(fileSourceName="novcc.txt", fileSourcePath="../", startTime=0, stopTime=10000, label=0)
abra.addDataFiles(fileSourceName="nowalk.txt", fileSourcePath="../", startTime=0, stopTime=10000, label=0)