def generateFirstOrderData(model, numIterations=10000, seqLength=5,
resets=True, suffix='train'):
print "Creating %d iteration file with seqLength %d" % (numIterations, seqLength)
print "Filename",
categoryList, initProbability, transitionTable = model
initProbability = initProbability.cumsum()
transitionTable = transitionTable.cumsum(axis=1)
outputFile = 'fo_%d_%d_%s.csv' % (numIterations, seqLength, suffix)
print "Filename", outputFile
fields = [('reset', 'int', 'R'), ('name', 'string', '')]
o = File(outputFile, fields)
seqIdx = 0
rand = numpy.random.rand()
catIdx = numpy.searchsorted(initProbability, rand)
for i in xrange(numIterations):
rand = numpy.random.rand()
if seqIdx == 0 and resets:
catIdx = numpy.searchsorted(initProbability, rand)
reset = 1
else:
catIdx = numpy.searchsorted(transitionTable[catIdx], rand)
reset = 0
o.write([reset,categoryList[catIdx]])
seqIdx = (seqIdx+1)%seqLength
o.close()
def makeDataset():
"""
"""
inputFile = 'numenta_air_Con.csv'
fields = [('gym', 'string', 'S'), ('address', 'string', ''),
('timestamp', 'datetime', 'T'), ('consumption', 'float', '')]
gymName = None
missing = 0
total = 0
# Create a the output file by parsing the customer given csv
with File('./hotgym2.csv', fields) as o:
with open(inputFile) as f:
# Skip header
f.readline()
# iterate over all the lines in the input file
for line in f.xreadlines():
# Parse the fields in the current line
record = _parseLine(line)
# Write the merged record to the output file
o.write(record)
if record[0] != gymName:
gymName = record[0]
print gymName
return total, missing
def makeDataset():
"""
"""
clubs = processAttendanceFiles()
clubs = processConsumptionFiles(clubs)
fields = [
('gym', 'string', 'S'),
('timestamp', 'datetime', 'T'),
('attendeeCount', 'int', ''),
('consumption', 'float', ''),
]
with File('gym.csv', fields) as f:
## write header
#f.write('Gym Name,Date,Time,Attendee Count,Consumption (KWH)\n')
for c in clubs.values():
for k, r in sorted(c.records.iteritems(),
key=operator.itemgetter(0)):
#dd = r.date[2]
#mm = r.date[1]
#yyyy = r.date[0]
#line = ','.join(str(x) for x in
# (c.name, '%d-%s-%d' % (dd, mmm, yyyy), r.time, r.attendeeCount, r.consumption))
#f.write(line + '\n')
f.write([r.club, r.timestamp, r.attendeeCount, r.consumption])
def generateFirstOrderData(model,
numIterations=10000,
seqLength=5,
resets=True,
suffix='train'):
print("Creating %d iteration file with seqLength %d" %
(numIterations, seqLength))
print("Filename", end=' ')
categoryList, initProbability, transitionTable = model
initProbability = initProbability.cumsum()
transitionTable = transitionTable.cumsum(axis=1)
outputFile = 'fo_%d_%d_%s.csv' % (numIterations, seqLength, suffix)
print("Filename", outputFile)
fields = [('reset', 'int', 'R'), ('name', 'string', '')]
o = File(outputFile, fields)
seqIdx = 0
rand = numpy.random.rand()
catIdx = numpy.searchsorted(initProbability, rand)
for i in range(numIterations):
rand = numpy.random.rand()
if seqIdx == 0 and resets:
catIdx = numpy.searchsorted(initProbability, rand)
reset = 1
else:
catIdx = numpy.searchsorted(transitionTable[catIdx], rand)
reset = 0
o.write([reset, categoryList[catIdx]])
seqIdx = (seqIdx + 1) % seqLength
o.close()
def _generateFile(filename, data):
"""
Parameters:
----------------------------------------------------------------
filename: name of .csv file to generate
"""
# Create the file
print "Creating %s..." % (filename)
numRecords, numFields = data.shape
fields = [('field%d'%(i+1), 'float', '') for i in range(numFields)]
outFile = File(filename, fields)
for i in xrange(numRecords):
outFile.write(data[i].tolist())
outFile.close()
