def catboost_train_model(model, trainx, trainy, cate_threshold=10):
model = clone(model)
cates = detect_cates_for_narrayx(trainx, cate_threshold)
trainx = transform_float_to_int_for_narrayx(trainx, cates)
model.fit(trainx, trainy, cates)
model.cates = cates
return model
def catboost_train(model,
trainx,
trainy,
testx,
cate_threshold=10,
predict_proba=True,
eval_testx=None,
cates=None):
model = catboost_train_model(model, trainx, trainy, cate_threshold, cates)
if (cates is None):
testx = transform_float_to_int_for_narrayx(testx, model.cates)
if (eval_testx is None):
return predict(model, "catboost", testx, predict_proba)
else:
return predict(model, "catboost", testx,
predict_proba), predict(model, "catboost", eval_testx,
predict_proba)
def predict(trained_model, model_name, testx, predict_proba=True):
if (model_name == "lgb"):
iteration = trained_model.best_iteration
if (iteration <= 0):
iteration = trained_model.current_iteration()
if (iteration < 0):
print("please set iteration, now it is {}".format(iteration))
exit(1)
if (predict_proba):
return trained_model.predict(testx, iteration)
else:
if(model_name == "catboost"):
testx = transform_float_to_int_for_narrayx(testx, trained_model.cates)
if (predict_proba):
return trained_model.predict_proba(testx)
else:
return trained_model.predict(testx)
def catboost_train(model, trainx, trainy, testx, cate_threshold=10, predict_proba=True):
model = catboost_train_model(model, trainx, trainy, cate_threshold)
testx = transform_float_to_int_for_narrayx(testx, model.cates)
return predict(model, "catboost", testx, predict_proba)