def svm_training(train_percent):
train_set_x,train_set_y,predict_set_x,target_y = data_input.getdata("processed.cleveland.data",train_percent)
clf = svm.SVC(gamma='scale')
clf.fit(train_set_x,np.ravel(train_set_y))
predict_set_y = clf.predict(predict_set_x)
dataframe = pd.DataFrame({'id':range(len(predict_set_y)),'predict_set_y':np.ravel(predict_set_y),'target_y':np.ravel(target_y)})
dataframe.to_csv("./svm/"+str(train_percent)+".csv",index=False,sep=',')
def GMM(train_percent):
train_set_x, train_set_y, predict_set_x, target_y = data_input.getdata(
"processed.cleveland.data", train_percent)
train_set_y = ravel(train_set_y)
target_y = ravel(target_y)
K = len(set(train_set_y))
Y = GaussianMixture(n_components=K,
covariance_type='full').fit(train_set_x)
Y = Y.predict(predict_set_x)
label = [0, 0, 0, 0, 0]
for k in range(K):
print target_y[Y == k]
def BR(train_percent):
from sklearn.linear_model import BayesianRidge
train_set_x, train_set_y, predict_set_x, target_y = data_input.getdata(
"processed.cleveland.data", train_percent)
train_set_y = normalize(train_set_y)
target_y = normalize(target_y)
clf = BayesianRidge(compute_score=True)
clf.fit(train_set_x, np.ravel(train_set_y))
# predict_set_y = np.round(clf.predict(predict_set_x))
predict_set_y0 = clf.predict(predict_set_x)
predict_set_y = []
for y in predict_set_y0:
predict_set_y.append(int(y))
dataframe = pd.DataFrame({
'id': range(len(predict_set_y)),
'predict_set_y': np.ravel(predict_set_y),
'target_y': np.ravel(target_y)
})
print np.sum(np.abs(np.ravel(np.round(predict_set_y)) - target_y))
dataframe.to_csv("./regression/BR" + str(train_percent) + ".csv",
index=False,
sep=',')
def LASSO(train_percent):
train_set_x, train_set_y, predict_set_x, target_y = data_input.getdata(
"processed.cleveland.data", train_percent)
alphas = np.linspace(0.5, 1, 100)
best_alp = 1
best_score = 10.0
for alp in alphas:
LASSO = Lasso(alpha=alp, fit_intercept=False, max_iter=1000)
LASSO.fit(train_set_x, train_set_y)
score = LASSO.score(predict_set_x, target_y)
if (score - 1.0)**2 < (best_score - 1.0)**2:
best_alp = alp
best_score = score
LASSO = Lasso(alpha=best_alp, fit_intercept=False, max_iter=1000)
LASSO.fit(train_set_x, train_set_y)
predict_set_y = np.round(LASSO.predict(predict_set_x))
dataframe = pd.DataFrame({
'id': range(len(predict_set_y)),
'predict_set_y': np.ravel(predict_set_y),
'target_y': np.ravel(target_y)
})
dataframe.to_csv("./regression/LASSO" + str(train_percent) + ".csv",
index=False,
sep=',')