app.py

import gradio as gr
import yfinance as yf
from prophet import Prophet
from sklearn.linear_model import LinearRegression, BayesianRidge
from sklearn.svm import SVR
from sklearn.preprocessing import MinMaxScaler
from statsmodels.tsa.arima.model import ARIMA
#from xgboost import XGBRegressor
import pandas as pd
import numpy as np
from datetime import datetime
import plotly.graph_objects as go
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense

def download_data(ticker, start_date='2010-01-01'):
    """
    주식 데이터를 다운로드하고 포맷을 조정하는 함수
    """
    data = yf.download(ticker, start=start_date)
    if data.empty:
        raise ValueError(f"No data returned for {ticker}")
    data.reset_index(inplace=True)
    if 'Adj Close' in data.columns:
        data = data[['Date', 'Adj Close']]
        data.rename(columns={'Date': 'ds', 'Adj Close': 'y'}, inplace=True)
    else:
        raise ValueError("Expected 'Adj Close' in columns")
    return data

def predict_future_prices(ticker, periods=1825):
    data = download_data(ticker)
    
    # Prophet 모델 생성 및 학습
    model_prophet = Prophet(daily_seasonality=False, weekly_seasonality=False, yearly_seasonality=True)
    model_prophet.fit(data)
    
    # 미래 데이터 프레임 생성 및 예측
    future = model_prophet.make_future_dataframe(periods=periods, freq='D')
    forecast_prophet = model_prophet.predict(future)
    
    # Linear Regression 모델 생성 및 학습
    model_lr = LinearRegression()
    X = pd.Series(range(len(data))).values.reshape(-1, 1)
    y = data['y'].values
    model_lr.fit(X, y)
    
    # 미래 데이터 프레임 생성 및 예측
    future_dates = pd.date_range(start=data['ds'].iloc[-1], periods=periods+1, freq='D')[1:].strftime('%Y-%m-%d')
    X_future = pd.Series(range(len(data), len(data) + len(future_dates))).values.reshape(-1, 1)
    future_lr = pd.DataFrame({'ds': future_dates, 'yhat': model_lr.predict(X_future)})
    
    
    # ARIMA 모델 생성 및 학습
    model_arima = ARIMA(data['y'], order=(1, 1, 1))
    model_arima_fit = model_arima.fit()
    forecast_arima = model_arima_fit.forecast(steps=periods)
    future_arima = pd.DataFrame({'ds': future_dates, 'yhat': forecast_arima})
    

    # LSTM 모델 생성 및 학습
    scaler = MinMaxScaler(feature_range=(0, 1))
    scaled_data = scaler.fit_transform(data['y'].values.reshape(-1, 1))
    X_train, y_train = [], []
    for i in range(60, len(scaled_data)):
        X_train.append(scaled_data[i-60:i, 0])
        y_train.append(scaled_data[i, 0])
    X_train, y_train = np.array(X_train), np.array(y_train)
    X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1))

    model_lstm = Sequential()
    model_lstm.add(LSTM(units=50, return_sequences=True, input_shape=(X_train.shape[1], 1)))
    model_lstm.add(LSTM(units=50))
    model_lstm.add(Dense(1))
    model_lstm.compile(loss='mean_squared_error', optimizer='adam')
    model_lstm.fit(X_train, y_train, epochs=10, batch_size=32)

    pred_lstm = []
    last_60_days = scaled_data[-60:]
    for i in range(periods):
        X_test = last_60_days.reshape(1, 60, 1)
        pred = model_lstm.predict(X_test)
        last_60_days = np.append(last_60_days[1:], pred)
        pred_lstm.append(pred[0, 0])

    pred_lstm = scaler.inverse_transform(np.array(pred_lstm).reshape(-1, 1))
    future_lstm = pd.DataFrame({'ds': future_dates[:len(pred_lstm)], 'yhat': pred_lstm.flatten()})
    
#  # XGBoost 모델 생성 및 학습
#   model_xgb = XGBRegressor(n_estimators=100, learning_rate=0.1)
#   model_xgb.fit(X.reshape(-1, 1), y)
#   future_xgb = pd.DataFrame({'ds': future_dates, 'yhat': model_xgb.predict(X_future)})
    
#    # SVR 모델 생성 및 학습
#   model_svr = SVR(kernel='rbf', C=1e3, gamma=0.1)
#   model_svr.fit(X.reshape(-1, 1), y)
#   future_svr = pd.DataFrame({'ds': future_dates, 'yhat': model_svr.predict(X_future)})
    
#    # Bayesian Regression 모델 생성 및 학습
#   model_bayes = BayesianRidge()
#   model_bayes.fit(X.reshape(-1, 1), y)
#   future_bayes = pd.DataFrame({'ds': future_dates, 'yhat': model_bayes.predict(X_future)})
    

    # 예측 결과 그래프 생성
    forecast_prophet['ds'] = forecast_prophet['ds'].dt.strftime('%Y-%m-%d')
    fig = go.Figure()
    fig.add_trace(go.Scatter(x=forecast_prophet['ds'], y=forecast_prophet['yhat'], mode='lines', name='Prophet Forecast (Blue)'))
    fig.add_trace(go.Scatter(x=future_lr['ds'], y=future_lr['yhat'], mode='lines', name='Linear Regression Forecast (Red)', line=dict(color='red')))
    fig.add_trace(go.Scatter(x=future_arima['ds'], y=future_arima['yhat'], mode='lines', name='ARIMA Forecast (Green)', line=dict(color='green')))
    fig.add_trace(go.Scatter(x=future_lstm['ds'], y=future_lstm['yhat'], mode='lines', name='LSTM Forecast (Orange)', line=dict(color='orange')))
#   fig.add_trace(go.Scatter(x=future_xgb['ds'], y=future_xgb['yhat'], mode='lines', name='XGBoost Forecast (Purple)', line=dict(color='purple')))
#   fig.add_trace(go.Scatter(x=future_svr['ds'], y=future_svr['yhat'], mode='lines', name='SVR Forecast (Brown)', line=dict(color='brown')))
#   fig.add_trace(go.Scatter(x=future_bayes['ds'], y=future_bayes['yhat'], mode='lines', name='Bayesian Regression Forecast (Pink)', line=dict(color='pink')))
    fig.add_trace(go.Scatter(x=data['ds'], y=data['y'], mode='lines', name='Actual (Black)', line=dict(color='black')))
    
    return fig, forecast_prophet[['ds', 'yhat', 'yhat_lower', 'yhat_upper']], future_lr[['ds', 'yhat']], future_arima[['ds', 'yhat']], future_lstm[['ds', 'yhat']], future_xgb[['ds', 'yhat']], future_svr[['ds', 'yhat']], future_bayes[['ds', 'yhat']]

    
# Gradio 인터페이스 설정 및 실행
with gr.Blocks() as app:
    with gr.Row():
        ticker_input = gr.Textbox(value="AAPL", label="Enter Stock Ticker for Forecast")
        periods_input = gr.Number(value=1825, label="Forecast Period (days)")
        forecast_button = gr.Button("Generate Forecast")
    
    forecast_chart = gr.Plot(label="Forecast Chart")
    forecast_data_prophet = gr.Dataframe(label="Prophet Forecast Data")
    forecast_data_lr = gr.Dataframe(label="Linear Regression Forecast Data")
    forecast_data_arima = gr.Dataframe(label="ARIMA Forecast Data")
    forecast_data_lstm = gr.Dataframe(label="LSTM Forecast Data")
#   forecast_data_xgb = gr.Dataframe(label="XGBoost Forecast Data")
#   forecast_data_svr = gr.Dataframe(label="SVR Forecast Data")
#   forecast_data_bayes = gr.Dataframe(label="Bayesian Regression Forecast Data")
    
    forecast_button.click(
        fn=predict_future_prices,
        inputs=[ticker_input, periods_input],
        outputs=[forecast_chart, forecast_data_prophet, forecast_data_lr, forecast_data_arima, forecast_data_lstm] #,forecast_data_xgb, forecast_data_svr, forecast_data_bayes]
    )

app.launch()