def XGBoost():
try:
params = {
'objective': 'binary:logistic',
'eta': 0.08,
'colsample_bytree': 0.886,
'min_child_weight': 1.1,
'max_depth': 7,
'subsample': 0.886,
'gamma': 0.1,
'lambda': 10,
'verbose_eval': True,
'eval_metric': 'auc',
'scale_pos_weight': 6,
'seed': 201703,
'missing': -1
}
xgbtrain = xgb.DMatrix(X_train, y_train)
xgbtest = xgb.DMatrix(X_test)
model = xgb.train(params, xgbtrain, num_boost_round=200)
xgb.save_model('xgb_time.model')
sys.exit(0)
predicted = model.predict(xgbtest)
return predicted
except:
print('die')
def update(self, score, xgb):
params = xgb.get_params()
if self.best_score > score:
prefix = str(-1 * score)[2:6]
self.best_score = score
self.best_params = params
self.best_model = xgb
self._write_json(self.best_params,
"tmp/best_params_{}.json".format(prefix))
xgb.save_model("tmp/best_model_{}.xgb".format(prefix))
print("best model updated: score: {}, params: {}".format(
score, params))
else:
pass
def run_cv(x_train, x_test, y_train, y_test):
x_train = x_train
conf.xgb_config()
tic = time.time()
data_message = 'X_train.shape={}, X_test.shape = {}'.format(
np.shape(x_train), np.shape(x_test))
print(data_message)
xgb = XGBooster(conf)
best_auc, best_round, cv_rounds, best_model = xgb.fit(x_train, y_train)
print('Training time cost {}s'.format(time.time() - tic))
xgb.save_model()
result_message = 'best_auc = {}, best_round = {}'.format(
best_auc, best_round)
print(result_message)
# now = time.strftime('%Y-%m-%d %H:%M')
result_saved_path = '../result/result_{}-{:.4f}.csv'.format(now, best_auc)
xgb_predict(best_model, x_test, y_test, save_result_path=result_saved_path)
'learning_rate':0.05,
'seed':2017,
'nthread':12,
'silent': 1
}
#plst+=[('eval_metric','auc')]
#evallist=[(x_val,'eval'),(x_train,'train')]
num_round=3000
plst=list(params.items())
plst+= [('eval_metric', 'auc')]
evallist = [(xgb_val, 'eval'), (xgb_train, 'train')]
xgb=xgb.train(params,xgb_train,num_boost_round=num_round)
print("save model")
xgb.save_model('./model/model4.txt')
print('开始预测')
preds_sub=xgb.predict(xgb_test)
test_tobepredicted['Ki']=preds_sub
test_tobepredicted.to_csv('./result/result4.csv',index=False)
# with open("result1.csv", "w") as f:
# sys.stdout = f
# print "Protein_ID,Molecule_ID"
# for index, Protein_ID in enumerate(user_ids):
# print "{},{}".format(userid, y[index])