prediction.py 3.3 KB
Newer Older
1
2
3
4
5
6
7
from config import DB
import datetime
from dateutil.relativedelta import relativedelta
import json
import os
import psycopg2
import sys
박예은's avatar
박예은 committed
8
9
10
11
import pandas as pd
import tensorflow as tf
import numpy as np

12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72


if __name__ == "__main__":

    dbconfig = {"host": DB["host"], "port": DB["port"], "user": DB["user"],
                "password": DB["password"], "database": DB["database"]}
    user_email = sys.argv[1]

    data_dir = os.getcwd() + "/src/data_processing/temp.csv"
    model_dir = os.getcwd() + "/src/data_processing/model.h5"

    # DB Connect and Make Cursor
    connection = psycopg2.connect(
        dbname=dbconfig["database"], user=dbconfig["user"], password=dbconfig["password"], host=dbconfig["host"], port=dbconfig["port"])
    cursor = connection.cursor()

    # Get Model and Params, then Make File and Variable
    cursor.execute(
        "SELECT model_file, params FROM \"Data_Processings\" WHERE host=%s", (user_email,))
    model_params = cursor.fetchone()

    blob_model = model_params[0]
    model_file = open(model_dir, "wb")
    model_file.write(blob_model)
    model_file.close()

    params = model_params[1]
    mean = json.loads(params["mean"])
    std = json.loads(params["std"])

    # Get User Info
    cursor.execute(
        "SELECT using_aircon, loc_code FROM \"Users\" WHERE email=%s", (user_email,))
    user_info = cursor.fetchone()

    # Get Weather Data and Make File
    today = datetime.date.today()
    yesterday = today - relativedelta(days=1)
    f_yesterday = "{0}-{1}-{2}".format(yesterday.year,
                                       yesterday.month, yesterday.day)

    cursor.execute(
        "SELECT collected_at as \"date\", temp as temp_out, humi as humi_out, press, wind_speed "
        + "From \"Weather_Outs\" "
        + "WHERE loc_code = %s AND collected_at >= %s", (user_info[1], f_yesterday,)
    )
    results = cursor.fetchall()

    data_file = open(data_dir, 'w')

    # header
    data_file.write("date,temp_out,humi_out,press,wind_speed\n")

    # contents
    for result in results:
        data_file.write("{0},{1},{2},{3},{4}\n".format(
            result[0], result[1], result[2], result[3], result[4]))

    data_file.close()
    cursor.close()

박예은's avatar
박예은 committed
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
    data_list = pd.read_csv(data_dir)
    new_data = data_list[-6:]

    date_time = pd.to_datetime(new_data['date'], format='%Y-%m-%d %H:%M')
    timestamp_s = date_time.map(datetime.datetime.timestamp)

    new_data = new_data[['temp_out', 'humi_out', 'press', 'wind_speed']]
    day = 24*60*60
    year = (365.2425)*day

    new_data['Day sin'] = np.sin(timestamp_s * (2 * np.pi / day))
    new_data['Day cos'] = np.cos(timestamp_s * (2 * np.pi / day))
    new_data['Year sin'] = np.sin(timestamp_s * (2 * np.pi / year))
    new_data['Year cos'] = np.cos(timestamp_s * (2 * np.pi / year))

    feature_cols = ['temp_out', 'humi_out', 'press',
                    'wind_speed', 'Day sin', 'Day cos', 'Year sin', 'Year cos']
    for col in feature_cols:
        new_data[col] =  (new_data[col] - mean[col]) / std[col]

    model_pro = tf.keras.models.load_model(model_dir)
    prediction = model_pro.predict(new_data)
    prediction = prediction * std['temp_out'] + mean['temp_out']

97
98
99
100
101
102
103
104
105

    # 사용한 파일 삭제
    if os.path.isfile(data_dir):
        os.remove(data_dir)

    if os.path.isfile(model_dir):
        os.remove(model_dir)

    print(prediction)