# binary_model = LinearSVC(random_state=0)
# binary_model = GaussianNB()
# binary_model = BernoulliNB()
binary_model = LogisticRegression()
# model = HOMER(base_clf=OneVsRestClassifier(binary_model, n_jobs=3),
# k=3,
# max_iter=20,
# # random_state=123456,
# # verbose=True,
# verbose=False)
# model = OneVsRestClassifier(binary_model, n_jobs=-1)
models = [# ClassifierChain(binary_model, n_jobs=5, verbose=2),
OneVsRestClassifier(binary_model, n_jobs=-1)
]
print "%d samples, %d features of training set:" % train_X.shape
print label_summary(train_y)
for model in models:
print "Using model: ", model
print "#" * 20
run_experiment(model, train_X, train_y, test_X, test_y)
print
# sample subset of all the data
rng = np.random.RandomState(0)
# SAMPLE_N = None
SAMPLE_N = 10000
if SAMPLE_N:
print "Sample size: %d" % SAMPLE_N
rows = rng.permutation(X.shape[0])[:SAMPLE_N]
X = X[rows, :]
y = y[rows, :]
else:
print "Sample size: all data"
SAMPLE_N = X.shape[0]
# sample train and test
train_ratio = 0.9
train_n = int(SAMPLE_N * train_ratio)
rows = rng.permutation(SAMPLE_N)
train_rows = rows[:train_n]
test_rows = rows[train_n:]
train_X = X[train_rows, :]
train_y = y[train_rows, :]
test_X = X[test_rows, :]
test_y = y[test_rows, :]
from exp_util import run_experiment
run_experiment(model, train_X, train_y, test_X, test_y, label_names)
