🧩Tensorflow Certification 취득하기 - part 11. 실전 ( Weekly US Retail)

Weekly US Retail

• Sequence (시계열) 데이터 다루기

Build and train a neural network to predict the time indexed variable of the univariate US diesel prices (On - Highway) All types for the period of 1994 - 2021.
Using a window of past 10 observations of 1 feature , train the model to predict the next 10 observations of that feature.
HINT: If you follow all the rules mentioned above and throughout this
question while training your neural network, there is a possibility that a
validation MAE of approximately 0.02 or less on the normalized validation
dataset may fetch you top marks.

Solution

순서 요약

1. import: 필요한 모듈 import
2. 전처리: 학습에 필요한 데이터 전처리를 수행합니다.
3. 모델링(model): 모델을 정의합니다.
4. 컴파일(compile): 모델을 생성합니다.
5. 학습 (fit): 모델을 학습시킵니다.

1. import 하기

필요한 모듈을 import 합니다.

import urllib
import pandas as pd
import tensorflow as tf
from tensorflow.keras.layers import Dense, Conv1D, LSTM, Bidirectional
from tensorflow.keras.models import Sequential
from tensorflow.keras.callbacks import ModelCheckpoint

# 아래 2줄 코드는 넣지 말아 주세요!!!
url = 'https://www.dropbox.com/s/eduk281didil1km/Weekly_U.S.Diesel_Retail_Prices.csv?dl=1'
urllib.request.urlretrieve(url, 'Weekly_U.S.Diesel_Retail_Prices.csv')

2.1 전처리 (Normalization)

# This function normalizes the dataset using min max scaling.
# DO NOT CHANGE THIS CODE
def normalize_series(data, min, max):
    data = data - min
    data = data / max
    return data

download_and_extract_data()

df = pd.read_csv('household_power_consumption.csv', sep=',', infer_datetime_format=True, index_col='datetime', header=0)
df.head(10)

2.2 전처리 (Windowed Dataset 생성)

# DO NOT CHANGE THIS.
def windowed_dataset(series, batch_size, n_past=10, n_future=10, shift=1):
    ds = tf.data.Dataset.from_tensor_slices(series)
    ds = ds.window(size=n_past + n_future, shift=shift, drop_remainder=True)
    ds = ds.flat_map(lambda w: w.batch(n_past + n_future))
    ds = ds.map(lambda w: (w[:n_past], w[n_past:]))
    return ds.batch(batch_size).prefetch(1)

2.3 전처리 (Dataset 로드)

df = pd.read_csv('Weekly_U.S.Diesel_Retail_Prices.csv', infer_datetime_format=True, index_col='Week of', header=0)
df.head(20)

# 특성 정의
N_FEATURES = len(df.columns)
N_FEATURES

2.4 전처리 (정규화 및 데이터 분할)

# 정규화 코드
data = df.values
data = normalize_series(data, data.min(axis=0), data.max(axis=0))

# 데이터 분할
SPLIT_TIME = int(len(data) * 0.8) # DO NOT CHANGE THIS
x_train = data[:SPLIT_TIME]
x_valid = data[SPLIT_TIME:]

BATCH_SIZE = 32  # 배치사이즈
N_PAST = 10      # 과거 데이터 (X)
N_FUTURE = 10    # 미래 데이터 (Y)
SHIFT = 1        # SHIFT

2.5 전처리 (train / validation set 구성)

train_set = windowed_dataset(series=x_train, batch_size=BATCH_SIZE,
                             n_past=N_PAST, n_future=N_FUTURE,
                             shift=SHIFT)

valid_set = windowed_dataset(series=x_valid, batch_size=BATCH_SIZE,
                             n_past=N_PAST, n_future=N_FUTURE,
                             shift=SHIFT)

3. 모델 정의 (Sequential)

이제 Modeling을 할 차례입니다.

model = tf.keras.models.Sequential([
    Conv1D(filters=32, kernel_size=5, padding='causal', activation='relu', input_shape=[N_PAST, 1]),
    Bidirectional(LSTM(32, return_sequences=True)),
    Bidirectional(LSTM(32, return_sequences=True)),
    Dense(32, activation='relu'),
    Dense(16, activation='relu'),
    Dense(N_FEATURES)
])

모델 결과 요약

model.summary()

4. 컴파일 (compile)

optimizer = tf.keras.optimizers.Adam(0.0001)
model.compile(optimizer=optimizer, loss=tf.keras.losses.Huber(), metrics=['mae'])

ModelCheckpoint: 체크포인트 생성

checkpoint_path = 'model/my_checkpoint.ckpt'
checkpoint = ModelCheckpoint(filepath=checkpoint_path,
                             save_weights_only=True,
                             save_best_only=True,
                             monitor='val_mae',
                             verbose=1)

5. 학습 (fit)

model.fit(train_set,
          validation_data=(valid_set),
          epochs=100,
          callbacks=[checkpoint])

학습 완료 후 Load Weights (ModelCheckpoint)

학습이 완료된 후에는 반드시 load_weights를 해주어야 합니다.
그렇지 않으면, 열심히 ModelCheckpoint를 만든 의미가 없습니다.

model.load_weights(checkpoint_path)