Upload 6 files

function/dr_prediction_deployment.py

@@ -0,0 +1,119 @@
+import sys
+import json
+import requests
+
+class DataRobotPredictionError(Exception):
+    """Raised if there are issues getting predictions from DataRobot"""
+
+
+def make_datarobot_deployment_predictions(data, deployment_id):
+    """
+    Make predictions on data provided using DataRobot deployment_id provided.
+    See docs for details:
+         https://app.datarobot.com/docs-jp/predictions/api/dr-predapi.html
+
+    Parameters
+    ----------
+    data : str
+        If using CSV as input:
+        Feature1,Feature2
+        numeric_value,string
+
+        Or if using JSON as input:
+        [{"Feature1":numeric_value,"Feature2":"string"}]
+
+    deployment_id : str
+        The ID of the deployment to make predictions with.
+
+    Returns
+    -------
+    Response schema:
+        https://app.datarobot.com/docs-jp/predictions/api/dr-predapi.html#response-schema
+
+    Raises
+    ------
+    DataRobotPredictionError if there are issues getting predictions from DataRobot
+    """
+    # Set HTTP headers. The charset should match the contents of the file.
+    headers = {
+        # As default, we expect CSV as input data.
+        # Should you wish to supply JSON instead,
+        # comment out the line below and use the line after that instead:
+        'Content-Type': 'text/plain; charset=UTF-8',
+        # 'Content-Type': 'application/json; charset=UTF-8',
+
+        'Authorization': 'Bearer {}'.format('NjQwMDVmNGI0ZDQzZDFhYzI2YThmZDJiOnVZejljTXFNTXNoUnlKMStoUFhXSFdYMEZRck9lY3dobnEvRFZ1aVBHbVE9'),
+        'DataRobot-Key': '84f96e49-d400-ec9c-92fc-30fc6e9329d1',
+    }
+    API_URL = 'https://jppdemo.orm.datarobot.com/predApi/v1.0/deployments/{deployment_id}/predictions'
+    url = API_URL.format(deployment_id=deployment_id)
+
+    # Prediction Explanations:
+    # See the documentation for more information:
+    # https://app.datarobot.com/docs-jp/predictions/api/dr-predapi.html#request-pred-explanations
+    # Should you wish to include Prediction Explanations or Prediction Warnings in the result,
+    # Change the parameters below accordingly, and remove the comment from the params field below:
+
+    params = {
+        # If explanations are required, uncomment the line below
+        # 'maxExplanations': 3,
+        # 'thresholdHigh': 0.5,
+        # 'thresholdLow': 0.15,
+        # If text explanations are required, uncomment the line below.
+        # 'maxNgramExplanations': 'all',
+        # Uncomment this for Prediction Warnings, if enabled for your deployment.
+        # 'predictionWarningEnabled': 'true',
+    }
+    # Make API request for predictions
+    predictions_response = requests.post(
+        url,
+        data=data,
+        headers=headers,
+        # Prediction Explanations:
+        # Uncomment this to include explanations in your prediction
+        # params=params,
+    )
+    _raise_dataroboterror_for_status(predictions_response)
+    # Return a Python dict following the schema in the documentation
+    return predictions_response.json()
+
+
+def _raise_dataroboterror_for_status(response):
+    """Raise DataRobotPredictionError if the request fails along with the response returned"""
+    try:
+        response.raise_for_status()
+    except requests.exceptions.HTTPError:
+        err_msg = '{code} Error: {msg}'.format(
+            code=response.status_code, msg=response.text)
+        raise DataRobotPredictionError(err_msg)
+
+
+def main(filename, deployment_id):
+    """
+    Return an exit code on script completion or error. Codes > 0 are errors to the shell.
+    Also useful as a usage demonstration of
+    `make_datarobot_deployment_predictions(data, deployment_id)`
+    """
+    MAX_PREDICTION_FILE_SIZE_BYTES = 52428800  # 50 MB
+    if not filename:
+        print(
+            'Input file is required argument. '
+            'Usage: python datarobot-predict.py <input-file.csv>')
+        return 1
+    data = open(filename, 'rb').read()
+    data_size = sys.getsizeof(data)
+    if data_size >= MAX_PREDICTION_FILE_SIZE_BYTES:
+        print((
+            'Input file is too large: {} bytes. '
+            'Max allowed size is: {} bytes.'
+        ).format(data_size, MAX_PREDICTION_FILE_SIZE_BYTES))
+        return 1
+    try:
+        predictions = make_datarobot_deployment_predictions(data, deployment_id)
+    except DataRobotPredictionError as exc:
+        print(exc)
+        return 1
+    return predictions
+
+# def prediction_formatting_to_dataframe(prediction_json):
+#     prediction_df = pd.jason_normalize(prediction_json)[['rowId'

function/get_estat.py

@@ -0,0 +1,31 @@
+import pandas as pd
+import datetime
+
+
+def get_household_survey():
+    # e-Statにユーザー登録し、APIキーを取得しておくこと
+    # URL: https://www.e-stat.go.jp/api/
+    API_KEY = "ddc1349cf530bdee69ca6a7ad6c0e2301aeb0780"
+
+    # 取得年月の設定
+    latest_year = int(datetime.datetime.now().strftime('%Y'))
+    year_period = 5
+    years = list(range(latest_year, latest_year - year_period, -1))
+    months = range(1, 13)
+    periods = []
+    for y in years:
+        y = y * 1_000_000
+        for m in months:
+            ym = y + m * 100 + m
+            periods.append(str(ym))
+    periods = ("%2C").join(periods)
+
+    # データ取得
+    url = f"http://api.e-stat.go.jp/rest/3.0/app/getSimpleStatsData?cdTab=01&cdCat02=03&cdArea=00000&cdTime={periods}&appId={API_KEY}&lang=J&statsDataId=0003343671&metaGetFlg=Y&cntGetFlg=N&explanationGetFlg=Y&annotationGetFlg=Y&sectionHeaderFlg=1&replaceSpChars=0"
+    df = pd.read_csv(url, header=28)
+    return df
+
+
+if __name__ == "__main__":
+    df = get_household_survey()
+    df.to_csv("household_survey.csv", index=False)

function/get_fish_qty.py

@@ -0,0 +1,107 @@
+import codecs
+import io
+import random
+import requests
+import time
+from datetime import date, timedelta
+from tqdm import tqdm
+from typing import Generator, Tuple
+
+import numpy as np
+import pandas as pd
+
+
+def date_range(
+    start: date, stop: date, step: timedelta = timedelta(1)
+) -> Generator[date, None, None]:
+    """startからendまで日付をstep日ずつループさせるジェネレータ"""
+    current = start
+    while current < stop:
+        yield current
+        current += step
+
+
+def get_url(download_date: date) -> Tuple[str, str]:
+    """ダウンロードするURLと日付の文字列を返す"""
+    month = download_date.strftime("%Y%m")
+    day = download_date.strftime("%Y%m%d")
+    return (
+        f"https://www.shijou-nippo.metro.tokyo.lg.jp/SN/{month}/{day}/Sui/Sui_K1.csv",
+        day,
+    )
+
+
+def content_wrap(content):
+    """1行目にヘッダ行が来るまでスキップする"""
+    buffer = ""
+    first = True
+    for line in io.BytesIO(content):
+        line_str = codecs.decode(line, "shift-jis")
+        if first:
+            if "品名" in line_str:
+                first = False
+                buffer = line_str
+            else:
+                continue
+        else:
+            buffer += line_str
+    return io.StringIO(buffer)
+
+
+def to_numeric(x):
+    """文字列を数値に変換する"""
+    if isinstance(x, str):
+        return float(x)
+    else:
+        return x
+
+
+def get_fish_price_data(start_date: date, end_date: date, use_fish_list) -> pd.core.frame.DataFrame:
+    """
+    東京卸売市場からデータを引っ張ってくる
+
+    :param start_date: 開始日
+    :param end_date: 終了日
+    :return: あじの値段を結合したデータ
+    """
+    columns = ['date'] + [i + '_卸売数量計(kg)' for i in use_fish_list] + ['全卸売数量計(kg)']
+    fish_qty_df = pd.DataFrame(columns=columns)
+    
+    iterator = tqdm(
+        date_range(start_date, end_date), total=(end_date - start_date).days
+    )
+
+    for download_date in iterator:
+        url, day = get_url(download_date)
+        iterator.set_description(day)
+        response = requests.get(url)
+
+        # URLが存在しないとき
+        temp_df = pd.DataFrame([{'date':day}])
+        if response.status_code == 404:
+            continue
+        assert (
+            response.status_code == 200
+        ), f"Unexpected HTTP response. Please check the website {url}."
+
+        df = pd.read_csv(content_wrap(response.content))
+        
+        
+        for i in use_fish_list:
+            temp = df.loc[df["品名"] == i, ['卸売数量計']]
+            
+            # display(temp)
+            
+            if len(temp) == 0:
+                temp_df[f'{i}_卸売数量計(kg)'] = 0
+            
+            
+            temp_df[f'{i}_卸売数量計(kg)'] = temp['卸売数量計'].sum()
+        
+        all_qty = df[['卸売数量計']].dropna().values[-1][0]
+        
+        temp_df['全卸売数量計(kg)'] = all_qty
+                
+        fish_qty_df = pd.concat([fish_qty_df, temp_df])
+        time.sleep(max(0.5 + random.normalvariate(0, 0.3), 0.1))
+    return fish_qty_df

function/prediction_func.py

@@ -0,0 +1,119 @@
+import pandas as pd
+import numpy as np
+import gradio as gr
+import datetime
+from dateutil.relativedelta import relativedelta
+from function import get_fish_qty, get_estat, dr_prediction_deployment
+
+import yaml
+with open('config.yaml') as file:
+    config = yaml.safe_load(file.read())
+
+def create_prediction_data(fish_sell_ach, oil_price_url, fuel_procurement_cost_url):
+    oil_price_df = pd.read_excel(oil_price_url, header=5)
+    oil_price_df = oil_price_df.rename(columns={oil_price_df.columns[0]:'年'})
+    oil_price_df['年'] = oil_price_df['年'].interpolate(method='ffill')
+    oil_price_df['年月'] = oil_price_df['年'] + oil_price_df['月'].astype(str) + '月'
+    oil_price_df['年月'] = pd.to_datetime(oil_price_df['年月'], format='%Y年%m月')
+    oil_price_df['年月'] = oil_price_df['年月'].apply(lambda x:x+relativedelta(months=3))
+    oil_price_df['年月'] = oil_price_df['年月'].apply(lambda x:''.join(str(x).split('-'))[:6]).astype(int)
+    oil_price_df = oil_price_df.drop(['年', '月'], axis=1)
+    for i in oil_price_df.columns:
+        if i != '年月':
+            oil_price_df = oil_price_df.rename(columns={i:f'{i}_lag3'})
+            oil_price_df[f'{i}_lag3'] = oil_price_df[f'{i}_lag3'].shift(1)
+    
+    fuel_procurement_cost_df = pd.read_excel(fuel_procurement_cost_url, header=4)
+    fuel_procurement_cost_df = fuel_procurement_cost_df.iloc[:, 3:]
+    for i in fuel_procurement_cost_df.columns:
+        if '\n' in i:
+            fuel_procurement_cost_df = fuel_procurement_cost_df.rename(columns={i:i.replace('\n', '')})
+
+    fuel_procurement_cost_df['燃料費調整単価適用期間'] = fuel_procurement_cost_df['燃料費調整単価適用期間'].interpolate(method='ffill')
+    fuel_procurement_cost_df['燃料費調整単価適用期間'] = pd.to_datetime(fuel_procurement_cost_df['燃料費調整単価適用期間'],
+                                                             format='%Y年\n%m月').astype(str).apply(lambda x:''.join(x.split('-'))[:6]).astype(int)
+    
+    col_list=['するめいか_卸売数量計(kg)',
+         'いわし_卸売数量計(kg)',
+         'ぶり・わらさ_卸売数量計(kg)',
+         '冷さけ_卸売数量計(kg)', 
+         '塩さけ_卸売数量計(kg)',
+         'さけます類_卸売数量計(kg)',
+         '全卸売数量計(kg)']
+    
+    for shift_i in [7, 14, 21, 28]:
+        change_col_list = [f'{i}_lag{shift_i}' for i in col_list]
+        fish_sell_ach[change_col_list] = fish_sell_ach[col_list].shift(shift_i)
+    
+    fish_sell_ach['target_date'] = fish_sell_ach['date'].apply(lambda x:int((pd.to_datetime(str(x))+relativedelta(months=1)).strftime('%Y%m%d')))
+    fish_sell_ach['年月'] = fish_sell_ach['target_date'].astype(str).str[:6].astype(int)
+    
+    prediction_df = pd.merge(fish_sell_ach, 
+                         oil_price_df,
+                         on='年月',
+                         how='left')
+    
+    for kind in fuel_procurement_cost_df['種別'].unique():
+        temp_df = fuel_procurement_cost_df.loc[fuel_procurement_cost_df['種別']==kind].drop('種別', axis=1)
+        temp_df = temp_df.rename(columns={temp_df.columns[0]:'年月'})
+        for i in temp_df.columns:
+            if i != '年月':
+                temp_df = temp_df.rename(columns={i:f'{i}_{kind}_lag1'})
+                temp_df['年月'] = pd.to_datetime(temp_df['年月'], format='%Y%m')
+                temp_df['年月'] = temp_df['年月'].apply(lambda x:x+relativedelta(months=1))
+                temp_df['年月'] = temp_df['年月'].apply(lambda x:''.join(str(x).split('-'))[:6]).astype(int)
+        prediction_df = pd.merge(prediction_df,
+                                 temp_df,
+                                 on='年月')
+    prediction_df = prediction_df.rename(columns={'date':'forecast_point'})
+    
+    return prediction_df
+
+def prediction_to_dr(oil_price_url, fuel_procurement_cost_url):
+    today = datetime.datetime.now()
+    last_prediction_result = pd.read_csv('data/prediction_result.csv')
+    last_time_fish_arch = pd.read_csv('data/fish_sell_ach.csv')
+    if (str(last_prediction_result['forecast_point'].max()) == today.strftime('%Y%m%d'))|(str(last_time_fish_arch['date'].max()) == int(today.strftime('%Y%m%d'))):
+        pass
+
+    else:
+        start_date = pd.to_datetime(str(last_time_fish_arch['date'].max()))
+        end_date = pd.to_datetime(today + relativedelta(days=1))
+        use_fish_list = config['use_fish_list']
+        temp_sell_ach = get_fish_qty.get_fish_price_data(start_date, end_date, use_fish_list)
+        temp_sell_ach['date'] = temp_sell_ach['date'].astype(int)
+        if str(temp_sell_ach['date'].max()) != today.strftime('%Y%m%d'):
+            pass
+
+        else:
+            temp_sell_ach = pd.concat([last_time_fish_arch,
+                                       temp_sell_ach.loc[~temp_sell_ach['date'].isin(last_time_fish_arch['date'].unique())]])
+            temp_sell_ach.to_csv('data/fish_sell_ach.csv', index=False)
+            prediction_df = create_prediction_data(temp_sell_ach, 
+                                              oil_price_url,
+                                              fuel_procurement_cost_url)
+            prediction_df = prediction_df.loc[(prediction_df['forecast_point'].astype(int)>last_prediction_result['forecast_point'].max())
+                                              &
+                                              (prediction_df['forecast_point'].astype(int)<=int(today.strftime('%Y%m%d')))].reset_index(drop=True)
+
+            display(prediction_df)
+            DEPLOYMENT_ID = '640d791796a6a52d92c368a0'
+            prediction_df.to_csv('data/temp_prediction.csv', index=False)
+            prediction_json = dr_prediction_deployment.main('data/temp_prediction.csv', DEPLOYMENT_ID)
+
+            prediction_result = pd.DataFrame({ 
+                                                'target_date':prediction_df['target_date'],
+                                                'forecast_point':prediction_df['forecast_point'],
+                                                '電気代':pd.json_normalize(prediction_json['data'])['prediction']
+                                            })
+
+            prediction_result = pd.merge(prediction_df,
+                                 prediction_result,
+                                 on=['target_date', 'forecast_point'])
+
+            last_prediction_result = pd.concat([last_prediction_result,
+                                                prediction_result])
+
+            last_prediction_result.to_csv('data/prediction_result.csv', index=False)
+            
+    return last_prediction_result

function/train_modeling.py

@@ -0,0 +1,221 @@
+import pandas as pd
+import numpy as np
+import gradio as gr
+import datetime
+from dateutil.relativedelta import relativedelta
+import datarobot as dr
+from function import get_fish_qty, get_estat, dr_prediction_deployment
+
+import yaml
+with open('config.yaml') as file:
+    config = yaml.safe_load(file.read())
+
+def create_train_data():
+    # ターゲットを抽出
+    household_survey = get_estat.get_household_survey()
+    expence_df = pd.DataFrame({'年月':household_survey['時間軸（月次）'].unique()})
+    cate='3.1 電気代'
+    temp_df = household_survey.loc[household_survey['品目分類（2020年改定）'] == cate]
+    unit = temp_df['unit'].unique()[0]
+    temp_df = temp_df.rename(columns={'value':f'{cate}_({unit})'})
+    expence_df = pd.merge(expence_df,
+                          temp_df[['時間軸（月次）', f'{cate}_({unit})']].rename(columns={'時間軸（月次）':'年月'}),
+                          on='年月',
+                          how='left')
+    expence_df = expence_df.rename(columns={'3.1 電気代_(円)':'電気代'})
+    expence_df['年月'] = pd.to_datetime(expence_df['年月'], format='%Y年%m月').astype(str).apply(lambda x:''.join(x.split('-'))[:6]).astype(int)
+    
+    # 原油価格を抽出し作成
+    oil_price_df = pd.read_excel(config['oil_price_url'], header=5)
+    oil_price_df = oil_price_df.rename(columns={oil_price_df.columns[0]:'年'})
+    oil_price_df['年'] = oil_price_df['年'].interpolate(method='ffill')
+    oil_price_df['年月'] = oil_price_df['年'] + oil_price_df['月'].astype(str) + '月'
+    oil_price_df['年月'] = pd.to_datetime(oil_price_df['年月'], format='%Y年%m月').astype(str).apply(lambda x:''.join(x.split('-'))[:6]).astype(int)
+    
+    # 燃料調達価格のデータを作成
+    fuel_procurement_cost_df = pd.read_excel(config['fuel_procurement_cost_url'], header=4)
+    fuel_procurement_cost_df = fuel_procurement_cost_df.iloc[:, 3:]
+    for i in fuel_procurement_cost_df.columns:
+        if '\n' in i:
+            fuel_procurement_cost_df = fuel_procurement_cost_df.rename(columns={i:i.replace('\n', '')})
+
+    fuel_procurement_cost_df['燃料費調整単価適用期間'] = fuel_procurement_cost_df['燃料費調整単価適用期間'].interpolate(method='ffill')
+    fuel_procurement_cost_df['燃料費調整単価適用期間'] = pd.to_datetime(fuel_procurement_cost_df['燃料費調整単価適用期間'],
+                                                             format='%Y年\n%m月').astype(str).apply(lambda x:''.join(x.split('-'))[:6]).astype(int)
+    for kind in fuel_procurement_cost_df['種別'].unique():
+        temp_df = fuel_procurement_cost_df.loc[fuel_procurement_cost_df['種別']==kind].drop('種別', axis=1)
+        temp_df = temp_df.rename(columns={temp_df.columns[0]:'年月'})
+        for i in temp_df.columns:
+            if i != '年月':
+                temp_df = temp_df.rename(columns={i:f'{i}_{kind}_lag1'})
+                temp_df[f'{i}_{kind}_lag1'] = temp_df[f'{i}_{kind}_lag1'].shift(1)
+        expence_df = pd.merge(expence_df,
+                              temp_df,
+                              on='年月',
+                              how='left')
+        
+    # 各データを結合
+    oil_price_df[['ブレント_lag3', 'ドバイ_lag3', 'WTI_lag3', 'OPECバスケット_lag3']] = oil_price_df[['ブレント', 'ドバイ', 'WTI', 'OPECバスケット']].shift(3)
+    expence_df = pd.merge(expence_df,
+                          oil_price_df[['ブレント_lag3', 'ドバイ_lag3', 'WTI_lag3', 'OPECバスケット_lag3', '年月']],
+                          on='年月',
+                          how='left')
+    
+    # 魚の卸売りデータを読み込み
+    last_time_fish_arch = pd.read_csv('data/fish_sell_ach.csv')
+    start_date = pd.to_datetime(str(int(last_time_fish_arch['date'].max())))
+    today = datetime.datetime.now()
+    end_date = pd.to_datetime(today + relativedelta(days=1))
+    use_fish_list = config['use_fish_list']
+    temp_sell_ach = get_fish_qty.get_fish_price_data(start_date, end_date, use_fish_list)
+    temp_sell_ach['date'] = temp_sell_ach['date'].astype(int)
+    sell_ach = pd.concat([last_time_fish_arch,
+                                              temp_sell_ach.loc[~temp_sell_ach['date'].isin(last_time_fish_arch['date'].unique())]])
+    sell_ach.to_csv('data/fish_sell_ach.csv', index=False)
+    
+    # trainデータの作成
+    sell_ach['target_date'] = sell_ach['date'].apply(lambda x:int((pd.to_datetime(str(x))+relativedelta(months=1)).strftime('%Y%m%d')))
+    sell_ach['年月'] = sell_ach['target_date'].astype(str).str[:6].astype(int)
+    
+    col_list=['するめいか_卸売数量計(kg)',
+                 'いわし_卸売数量計(kg)',
+                 'ぶり・わらさ_卸売数量計(kg)',
+                 '冷さけ_卸売数量計(kg)', 
+                 '塩さけ_卸売数量計(kg)',
+                 'さけます類_卸売数量計(kg)',
+                 '全卸売数量計(kg)']
+    
+    for shift_i in [7, 14, 21, 28]:
+        change_col_list = [f'{i}_lag{shift_i}' for i in col_list]
+        sell_ach[change_col_list] = sell_ach[col_list].shift(shift_i)
+    
+    sell_ach = sell_ach.rename(columns={'date':'forecast_point'})
+    train_df = pd.merge(expence_df, 
+                        sell_ach,
+                        on='年月')
+    train_df.to_csv('data/train.csv', index=False)
+    
+    return train_df
+
+
+def modeling():
+    train_df = create_train_data()
+    # モデリングに必要な各設定値
+    ## データロボットとの接続設定
+    token = 'NjQwMDVmNGI0ZDQzZDFhYzI2YThmZDJiOnVZejljTXFNTXNoUnlKMStoUFhXSFdYMEZRck9lY3dobnEvRFZ1aVBHbVE9'
+    ### デモ環境これっぽい
+    endpoint = 'https://app.datarobot.com/api/v2' 
+
+    ## プロジェクト名
+    project_name = f'{datetime.datetime.now().strftime("%Y%m%d")}_ESTYLEU_電気代予測_再学習'
+
+    ## 各種設定
+    ### 特徴量設定
+    target = '電気代'
+    feature_timeline = 'target_date'    #時系列
+    not_use_feature = ['年月', 'forecast_point']
+    # 最適化指標
+    metric = 'RMSE'
+    ### ギャップ
+    gap='P0Y'       #　これで0？要確認
+    ### バックテストの数
+    number_of_backtests = 1
+    end_date = int(train_df[feature_timeline].max())
+    ### 日付
+    holdout_end_date=pd.to_datetime(str(end_date))
+    holdout_start_date=holdout_end_date - relativedelta(years=1)
+    backtest_end_date = holdout_start_date - relativedelta(days=1)
+    backtest_start_date = backtest_end_date - relativedelta(years=1)
+    train_end_date = backtest_start_date - relativedelta(days=1)
+    train_start_date = pd.to_datetime(str(int(train_df[feature_timeline].min())))
+
+    ### モデリングモード
+    # mode = dr.AUTOPILOT_MODE.QUICK
+    mode = dr.AUTOPILOT_MODE.FULL_AUTO
+    dr.Client(
+                    endpoint=endpoint,
+                    token=token
+                )
+
+    # バックテスト設定
+    backtests_setting = [dr.BacktestSpecification(
+                                                                            index=0,
+                                                                            primary_training_start_date=train_start_date, 
+                                                                            primary_training_end_date=train_end_date,
+                                                                            validation_start_date=backtest_start_date, 
+                                                                            validation_end_date=backtest_end_date
+                                                                        )]
+
+    spec = dr.DatetimePartitioningSpecification(
+                                                                        feature_timeline,
+                                                                        use_time_series=False,
+                                                                        disable_holdout=False,
+                                                                        holdout_start_date=holdout_start_date,
+                                                                        holdout_end_date=holdout_end_date,
+                                                                        gap_duration=gap,
+                                                                        number_of_backtests=number_of_backtests,
+                                                                        backtests=backtests_setting,
+                                                                    )
+
+    use_feature_list = train_df.columns.to_list()
+
+    print('now creating project')
+    project = dr.Project.create(
+                                                train_df, 
+                                                project_name=project_name
+                                            )
+
+    raw = [feat_list for feat_list in project.get_featurelists() if feat_list.name == 'Informative Features'][0]
+    raw_features = [feat for feat in raw.features if f'{feature_timeline} ' in feat]
+
+    for i in not_use_feature:
+        if i in use_feature_list:
+            use_feature_list.remove(i)
+
+    use_feature_list = use_feature_list.extend(raw_features)
+    print("start modeling")
+    project.analyze_and_model(
+                                                target = target,
+                                                mode = mode,
+                                                partitioning_method=spec,
+                                                max_wait=3000,
+                                                worker_count=-1,
+                                                featurelist_id = project.create_featurelist('モデリング', use_feature_list).id
+                                            )
+    project.wait_for_autopilot()
+    project.unlock_holdout()
+    
+    model_df = pd.DataFrame(
+        [[model.id,
+          model.model_type,
+          model.metrics['RMSE']['validation'],
+          model.metrics['RMSE']['backtesting'],
+          model.metrics['RMSE']['holdout'],
+          model] for model in project.get_datetime_models() if model.model_type != 'Baseline Predictions Using Most Recent Value'],
+        columns=['ID', 'モデル名', 'バックテスト1', '全てのバックテスト', 'holdout', 'model'])
+    model_df = model_df.sort_values('holdout').reset_index(drop=True)
+    
+    model = model_df['model'][0]
+    
+    try:
+        model_management_df = read_csv('data/model_management.csv')
+    except:
+        model_management_df = pd.DataFrame()
+        
+    temp_model_management_df = pd.DataFrame({
+                                                                           '作成日':[int(datetime.datetime.now().strftime('%Y%m%d'))],
+                                                                           '作成時間':[int(datetime.datetime.now().strftime('%H%M%S'))],
+                                                                           'project_url':[project.get_uri()],
+                                                                           'model_url':[model.get_uri()],
+                                                                           'model_type':[model.model_type]
+                                                                         })
+    
+    model_management_df = pd.concat([model_management_df,
+                                                            temp_model_management_df])
+    model_management_df.to_csv('data/model_management.csv')
+    
+    deployment = dr.Deployment.get(deployment_id='640d791796a6a52d92c368a0')
+    
+    deployment.replace_model(model.id, dr.enums.MODEL_REPLACEMENT_REASON.SCHEDULED_REFRESH)
+    
+