MINDIFF, PERSISTENCE)
print 'Using spambase data folded {} ways. Testing on fold {}.'.format(
opt['folds'], opt['testfold'])
# stopping conditions
if opt['rounds'] is None:
def doround(err, rndct):
return checkconverged(err)
else:
def doround(err, rndct):
return countdown(rndct)
# load from file
spambase.load()
# roll into folds
folds = [[] for i in xrange(opt['folds'])]
k = 0 # kurrent fold
for dp in spambase.data:
# change the 0,1 labels to -1,1
dp.label = 1 if dp.label else -1
# add to the current fold & switch to the next fold
folds[k].append(dp)
k = (k + 1) % opt['folds']
# unroll to testing & training
testing = folds.pop(opt['testfold'])
training = reduce(lambda acc, cur: acc + cur, folds)
del folds
def main():
spambase.load('spambase.data')
sc = StumpCollection(spambase.trainset)
auc = resultset.auc(roc)
with open('{}-{}_lambda={}_auc={}'.format(regression.__name__, gdname,
suffix, auc).lower(),
mode='wb') as fd:
for fpr, tpr in roc:
fd.write('{}, {}\n'.format(fpr, tpr))
###############################################################################
if __name__ == '__main__':
print
# load from file
spambase.load()
d = spambase.data
# zscore feature values
def pre(lon):
mu = stats.mean(lon)
sd = stats.stddev(stats.mvue(lon, mu))
return stats.zscore(lon, mu, sd)
dataset.applykernel(d, pre)
# insert phantom features
for dp in d:
dp.features = [1.0] + list(dp.features)
# roll into folds
folds = [[] for i in xrange(FOLDCOUNT)]
def main():
spambase.load('spambase.data')
sc = StumpCollection(spambase.trainset)
