def rf_final():
from sklearn import ensemble
tr_x, tr_y, _, __, te_x, te_y = load_data()
fl = open(RF_FILE)
pars = pickle.load(fl)
fl.close()
opt = reduce(lambda x, y: x if x[2] > y[2] else y, pars)
print 'RF best: ', opt[1]
m = ensemble.RandomForestClassifier(criterion='gini', **opt[1])
m = m.fit(tr_x, tr_y)
proba = m.predict_proba(te_x)
fpr, tpr, thresh = roc_curve(te_y, proba[:, 1])
fpr_tpr = np.vstack((fpr, tpr)).T
np.savetxt('rf_fpr_tpr.csv', fpr_tpr, delimiter=',')
def dt_final():
from sklearn import tree
tr_x, tr_y, _, __, te_x, te_y = load_data()
fl = open(DT_FILE)
pars = pickle.load(fl)
fl.close()
opt = reduce(lambda x, y: x if x[2] > y[2] else y, pars)
print 'DT best: ', opt[1]
m = tree.DecisionTreeClassifier(criterion='entropy', **opt[1])
m = m.fit(tr_x, tr_y)
proba = m.predict_proba(te_x)
fpr, tpr, thresh = roc_curve(te_y, proba[:, 1])
fpr_tpr = np.vstack((fpr, tpr)).T
np.savetxt('dt_fpr_tpr.csv', fpr_tpr, delimiter=',')