def __init__(self, filename):
"""Input: path and name of the file of a trace; how many filtering steps should be used for sliding window filtering"""
self.__id = int(os.path.basename(filename).split(".")[0])
x = []
y = []
with open(filename, "r") as trainfile:
trainfile.readline() # skip header
for line in trainfile:
items = line.split(",", 2)
x.append(float(items[0]))
y.append(float(items[1]))
triplength = distance(x[0], y[0], x[-1], y[-1])
self.triptime = len(x)
xvar, yvar = bettervariance(x, y)
overlaps = overlap(x, y)
self.featurelist = []
self.featurelist.append(triplength)
self.featurelist.append(self.triptime)
self.featurelist.append(xvar)
self.featurelist.append(yvar)
self.featurelist.append(overlaps)
v, distancecovered = velocities_and_distance_covered(x, y)
vfiltered = reject_outliers(v, 3)
maxspeed = max(vfiltered)
medianspeed = median(vfiltered)
meanspeed, varspeed = meanstdv(vfiltered)
angles, jumps = self.angles_and_jumps()
anglespeed = [speed * angle for (speed, angle) in zip(v, angles)]
anglespeedfiltered = reject_outliers(anglespeed, 3)
maxanglespeed = max(anglespeedfiltered)
mediananglespeed = median(anglespeedfiltered)
meananglespeed, varanglespeed = meanstdv(anglespeedfiltered)
totalangle = sum(angles)
maxangle = max(angles)
minangle = min(angles)
medianangle = max(angles)
acc, dec, stills = getratio(vfiltered, medianspeed)
if (acc + dec) == 0:
accratio = 0.5
else:
accratio = acc / (acc + dec)
self.featurelist.append(distancecovered)
self.featurelist.append(maxspeed)
self.featurelist.append(medianspeed)
self.featurelist.append(meanspeed)
self.featurelist.append(varspeed)
self.featurelist.append(maxanglespeed)
self.featurelist.append(mediananglespeed)
self.featurelist.append(meananglespeed)
self.featurelist.append(varanglespeed)
self.featurelist.append(totalangle)
self.featurelist.append(medianangle)
self.featurelist.append(maxangle)
self.featurelist.append(minangle)
self.featurelist.append(jumps)
self.featurelist.append(accratio)
self.featurelist.append(acc)
self.featurelist.append(dec)
self.featurelist.append(stills)
def analysis(trainfoldername, testfoldername, outdir, referencenum):
"""
Start the analysis
Input:
1) Path to the driver directory
2) Path where the submission file should be written
3) Number of drivers to compare against
"""
seed(42)
start = datetime.now()
submission_id = datetime.now().strftime("%H_%M_%B_%d_%Y")
trainfolders = [os.path.join(trainfoldername, f) for f in os.listdir(trainfoldername) if os.path.isdir(os.path.join(trainfoldername, f))]
referencefolders = [trainfolders[i] for i in sorted(sample(xrange(len(trainfolders)), referencenum))]
referencedrivers = []
for referencefolder in referencefolders:
referencedrivers.append(Driver(referencefolder))
generatedata(referencedrivers)
testdrivers = []
testfolders = [os.path.join(testfoldername, f) for f in os.listdir(testfoldername) if os.path.isdir(os.path.join(testfoldername, f))]
for testfolder in testfolders:
testdrivers.append(Driver(testfolder))
generatetestdata(testdrivers)
results = Parallel(n_jobs=10)(delayed(perform_analysis)(trainfolder) for trainfolder in trainfolders)
with open(os.path.join(outdir, "testing_results_{0}.txt".format(submission_id)), 'w') as writefile:
for item in results:
writefile.write("%.4f\n" % item)
mean = sum(results)/len(results)
writefile.write("Mean: %.4f\n" % mean)
writefile.write("Median: %.4f\n" % median(results))
writefile.write("Min: %.4f\n" % min(results))
print 'Done, elapsed time: %s' % str(datetime.now() - start)
def analysis(trainfoldername, testfoldername, outdir, referencenum):
"""
Start the analysis
Input:
1) Path to the driver directory
2) Path where the submission file should be written
3) Number of drivers to compare against
"""
seed(42)
start = datetime.now()
submission_id = datetime.now().strftime("%H_%M_%B_%d_%Y")
trainfolders = [
os.path.join(trainfoldername, f) for f in os.listdir(trainfoldername)
if os.path.isdir(os.path.join(trainfoldername, f))
]
referencefolders = [
trainfolders[i]
for i in sorted(sample(xrange(len(trainfolders)), referencenum))
]
referencedrivers = []
for referencefolder in referencefolders:
referencedrivers.append(Driver(referencefolder))
generatedata(referencedrivers)
testdrivers = []
testfolders = [
os.path.join(testfoldername, f) for f in os.listdir(testfoldername)
if os.path.isdir(os.path.join(testfoldername, f))
]
for testfolder in testfolders:
testdrivers.append(Driver(testfolder))
generatetestdata(testdrivers)
results = Parallel(n_jobs=10)(delayed(perform_analysis)(trainfolder)
for trainfolder in trainfolders)
with open(
os.path.join(outdir,
"testing_results_{0}.txt".format(submission_id)),
'w') as writefile:
for item in results:
writefile.write("%.4f\n" % item)
mean = sum(results) / len(results)
writefile.write("Mean: %.4f\n" % mean)
writefile.write("Median: %.4f\n" % median(results))
writefile.write("Min: %.4f\n" % min(results))
print 'Done, elapsed time: %s' % str(datetime.now() - start)
