def setAlgorithms(self, algos):
self.algos = {}
wsum = sum([algo[1] for algo in algos])
for algo in algos:
self.algos[algo[0]] = {
"weight": algo[1] * 1.0 / wsum,
"object": loadAlgorithm(algo[0])
}
def testAlgorithm(dataset, algorithm):
print "[%s] Splitting learn & probe" % (algorithm)
# (learn,probe) = dataset.splitInLearnAndProbe(lambda row:row[4]%80==0)
# (learn,probe) = dataset.splitInLearnAndProbe(lambda row:row[1]!=33 )
learn_ids = (
33,
80753,
65893,
51330,
41201,
24426,
23932,
16853,
6534,
2131,
43152,
43153,
43155,
43156,
43161,
43162,
43163,
43164,
43169,
)
(learn, probe) = dataset.splitInLearnAndProbe(lambda row: row[1] not in learn_ids)
# (learn,probe) = dataset.splitInLearnAndProbe(lambda row:row[1]!=32)
a = loadAlgorithm(algorithm)
learn._probe = probe
print "[%s] Training dataset (%s rows of %s total)..." % (
algorithm,
len(learn.data),
len(learn.data) + len(probe.data),
)
a.trainDataset(learn)
print "[%s] Computing RMSE..." % (algorithm)
# return (learn.rmse(a),probe.rmse(a))
return (0, probe.rmse(a))
def testAlgorithm(dataset, algorithm):
print "[%s] Splitting learn & probe" % (algorithm)
#(learn,probe) = dataset.splitInLearnAndProbe(lambda row:row[4]%80==0)
#(learn,probe) = dataset.splitInLearnAndProbe(lambda row:row[1]!=33 )
learn_ids = (33, 80753, 65893, 51330, 41201, 24426, 23932, 16853, 6534,
2131, 43152, 43153, 43155, 43156, 43161, 43162, 43163, 43164,
43169)
(learn,
probe) = dataset.splitInLearnAndProbe(lambda row: row[1] not in learn_ids)
#(learn,probe) = dataset.splitInLearnAndProbe(lambda row:row[1]!=32)
a = loadAlgorithm(algorithm)
learn._probe = probe
print "[%s] Training dataset (%s rows of %s total)..." % (
algorithm, len(learn.data), len(learn.data) + len(probe.data))
a.trainDataset(learn)
print "[%s] Computing RMSE..." % (algorithm)
#return (learn.rmse(a),probe.rmse(a))
return (0, probe.rmse(a))
def setAlgorithms(self,algos):
self.algos = {}
wsum = sum([algo[1] for algo in algos])
for algo in algos:
self.algos[algo[0]] = {"weight":algo[1]*1.0/wsum,"object":loadAlgorithm(algo[0])}
