def post_json_prediction_accuracy(company_code, company_name, start_date,
end_date):
while start_date <= end_date:
w1, w2, w3 = prediction.start(company_code, start_date)
# 바뀐 방식으로 예측 종가를 계산
predicted_value = closing_calculation.predict(company_code, start_date,
w1, w2, w3)
predicted_value = round(predicted_value / 10) * 10
lstm_price, closing_price = get_lstm_prediction_data(
company_code, start_date - datetime.timedelta(days=1))
_, correct_closing_price = get_lstm_prediction_data(
company_code, start_date)
result = {
"date": str(start_date),
"lstm_closing_price": float(lstm_price),
"correct_closing_price": float(correct_closing_price),
"predicted_closing_price": float(predicted_value),
"company_name": str(company_name),
"company_code": str(company_code),
}
# print(result)
# elasticsearch 로 데이터 전송
json_data = json.dumps(result, ensure_ascii=False)
index = f"prediction-accuracy-{start_date}"
store_record(index, json_data)
start_date += datetime.timedelta(days=1)
def post_json_weight(company_code, company_name, start_date, end_date):
while start_date <= end_date:
w1, w2, w3 = prediction.start(company_code, start_date)
result = {
"date": str(start_date),
"lstm_weight": float(w1),
"emotional_weight": float(w2),
"per_weight": float(w3),
"company_name": str(company_name),
"company_code": str(company_code),
}
# print(result)
# elasticsearch 로 데이터 전송
json_data = json.dumps(result, ensure_ascii=False)
index = f"prediction-weight-{start_date}"
store_record(index, json_data)
start_date += datetime.timedelta(days=1)
if __name__ == '__main__':
for company in COMPANIES:
# 1. PER 정보 크롤링
per_crawler.start(company[0])
# 2. 뉴스기사 크롤링
news_contents_crawler.start(company[0], START_DATE, END_DATE)
# 3. kosac 을 이용해 뉴스기사 데이터 전처리
kosac_preprocessor.start(company[0], START_DATE, END_DATE)
# 4. 뉴스기사 감성분석
news_contents_sentimental_analysis.start(company[0], START_DATE, END_DATE)
learning_date = START_DATE
while learning_date <= END_DATE:
# 5. lstm 계산
lstm_calculator.start(company[0], learning_date)
learning_date += datetime.timedelta(days=1)
# 6. 가중치 a, b, c 계산
w1, w2, w3 = prediction.start(company[0], learning_date)
# 7. 계산된 가중치 a, b, c 를 활용하여 다음날 주가 예측
predicted_value = closing_calculation.predict(company[0], learning_date, w1, w2, w3)
# 7. elasticsearch 로 데이터 전송
# 날짜를 START_DATE 와 END_DATE 를 같게 해야 합니다.
#elasticsearch_client.post_data(company[0], company[1], END_DATE, END_DATE)
help=f"which prediction method execute. {execution_help}",
required=True,
)
parser.add_argument(
"-o",
"--output",
dest="output",
type=str,
nargs=None,
help=f"name of the file that will have the final submission",
required=True,
)
parser.add_argument(
"-e",
"--experiment",
dest="experiment",
type=str,
nargs=None,
help="description of the experiment",
required=True,
)
args = parser.parse_args()
method = args.method
train_file = "./data/02_intermediate/opiniones_train_opiniones_1.csv"
test_file = "./data/02_intermediate/opiniones_test_opiniones_2.csv"
submission_file = f"./data/07_model_output/{args.output}.csv"
logger.info(f"Experiment description: {args.experiment}")
start(method, train_file, test_file, submission_file, logger)
def predict_number(sub_img):
"""calls prediction class to predict the number"""
prediction.start(sub_img)
random_state = cfg.getProperty("Challenge.LibroQueLeo.randomState")
if method == "ensamble":
# (predictor_identifier, weight), model_file
weights = cfg.getProperty("Challenge.LibroQueLeo.Ensamble.Weights")
models = cfg.getProperty("Challenge.LibroQueLeo.Ensamble.Models")
model_tuples = []
for predictor, weight in weights.items():
model_file = model_path + models[predictor]
model_tuples += [((predictor, weight), model_file)]
ensamble(
model_tuples,
train_file,
test_file,
submission_file,
logger,
random_state,
)
else:
start(
method,
train_file,
test_file,
submission_file,
logger,
model_file=model_file,
random_state=random_state,
)