TensorFlow 기초 16 - AutoMPG dataset으로 자동차 연비 예측 모델(표준화)

 

 

# AutoMPG dataset으로 자동차 연비 예측모델
import pandas as pd
import matplotlib.pyplot as plt 
import seaborn as sns
import tensorflow as tf
from keras import layers

dataset = pd.read_csv('../testdata/auto-mpg.csv', na_values='?') # na_values='?' na를 ?표로 대체함
print(dataset.head(2))
del dataset['car name']

print(dataset.corr())
dataset.drop(['acceleration', 'model year', 'origin'], axis='columns', inplace=True)
print(dataset.info())
print(dataset.corr())

# print(dataset.isna().sum())  # 6개 있다.
dataset = dataset.dropna()
# print(dataset.isna().sum())  # 없애기 성공

# 시각화 
# sns.pairplot(dataset[['mpg','displacement', 'horsepower', 'weight', 'acceleration']], diag_kind='kde')
# plt.show()

# train / test split
train_dataset = dataset.sample(frac=0.7, random_state=123)
test_dataset = dataset.drop(train_dataset.index)
print(train_dataset.shape, test_dataset.shape) # (274, 8) (118, 8)

# 표준화
train_stat = train_dataset.describe()
# print(train_stat)
train_stat.pop('mpg') # mpg는 label로 사용할 것이므로..
train_stat = train_stat.transpose()
print(train_stat)

train_label = train_dataset.pop('mpg') # mpg만 따로 뽑아봄
print(train_label[:2])
test_label = test_dataset.pop('mpg') # mpg만 따로 뽑아봄
print(test_label[:2])

def st_func(x):  # 표준화 함수
    return (x - train_stat['mean']) / train_stat['std']

# print(st_func(10))
# print(train_dataset[:2])
# print(st_func(train_dataset[:2]))
st_train_data = st_func(train_dataset)  # feature
st_test_data = st_func(test_dataset)    # feature
print(st_train_data.columns)

print()
# model
from keras.models import Sequential
from keras.layers import Dense

def build_model():
    network = Sequential([
        Dense(units=64, activation='relu', input_shape=[4]),
        Dense(units=64, activation='relu'),
        Dense(units=1, activation='linear')
        
    ])
    opti= tf.keras.optimizers.RMSprop(0.001)
    network.compile(optimizer=opti, loss='mean_squared_error', metrics=['mean_absolute_error', 'mean_squared_error'])
    
    return network

model = build_model()
print(model.summary())

# fit() 전에 모델을 실행해도 됨. 다만, 성능은 기대하지 않음.
# print(model.predict(st_train_data[:1]))

epochs = 10000
early_stop = tf.keras.callbacks.EarlyStopping(monitor='val_loss', mode='auto', patience=5)

history = model.fit(st_train_data, train_label, batch_size=32, epochs=epochs, 
                    validation_split=0.2, verbose=1, callbacks=[early_stop])

df = pd.DataFrame(history.history)
print(df)

def plot_history(history):
    hist = pd.DataFrame(history.history)
    hist['epoch'] = history.epoch
 
    plt.figure(figsize=(8,12))
    plt.subplot(2,1,1)
    plt.xlabel('Epoch')
    plt.ylabel('Mean Abs Error [MPG]')
    plt.plot(hist['epoch'], hist['mean_absolute_error'],label='Train Error')
    plt.plot(hist['epoch'], hist['val_mean_absolute_error'],label = 'Val Error')
    plt.legend()

    plt.subplot(2,1,2)
    plt.xlabel('Epoch')
    plt.ylabel('Mean Square Error [$MPG^2$]')
    plt.plot(hist['epoch'], hist['mean_squared_error'],label='Train Error')
    plt.plot(hist['epoch'], hist['val_mean_squared_error'],label = 'Val Error')
    plt.legend()
    plt.show()

plot_history(history)

# 모델 평가
loss, mae, mse = model.evaluate(st_test_data, test_label)
print('loss : {:5.3f}'.format(loss))
print('mae : {:5.3f}'.format(mae))
print('mse : {:5.3f}'.format(mse))

from sklearn.metrics import r2_score
print('설명력 : ', r2_score(test_label, model.predict(st_test_data))) # 0.6687


<console>
    mpg  cylinders  displacement  ...  model year  origin                   car name
0  18.0          8         307.0  ...          70       1  chevrolet chevelle malibu
1  15.0          8         350.0  ...          70       1          buick skylark 320

[2 rows x 9 columns]
                   mpg  cylinders  ...  model year    origin
mpg           1.000000  -0.775396  ...    0.579267  0.563450
cylinders    -0.775396   1.000000  ...   -0.348746 -0.562543
displacement -0.804203   0.950721  ...   -0.370164 -0.609409
horsepower   -0.778427   0.842983  ...   -0.416361 -0.455171
weight       -0.831741   0.896017  ...   -0.306564 -0.581024
acceleration  0.420289  -0.505419  ...    0.288137  0.205873
model year    0.579267  -0.348746  ...    1.000000  0.180662
origin        0.563450  -0.562543  ...    0.180662  1.000000

[8 rows x 8 columns]
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 398 entries, 0 to 397
Data columns (total 5 columns):
 #   Column        Non-Null Count  Dtype  
---  ------        --------------  -----  
 0   mpg           398 non-null    float64
 1   cylinders     398 non-null    int64  
 2   displacement  398 non-null    float64
 3   horsepower    392 non-null    float64
 4   weight        398 non-null    int64  
dtypes: float64(3), int64(2)
memory usage: 15.7 KB
None
                   mpg  cylinders  displacement  horsepower    weight
mpg           1.000000  -0.775396     -0.804203   -0.778427 -0.831741
cylinders    -0.775396   1.000000      0.950721    0.842983  0.896017
displacement -0.804203   0.950721      1.000000    0.897257  0.932824
horsepower   -0.778427   0.842983      0.897257    1.000000  0.864538
weight       -0.831741   0.896017      0.932824    0.864538  1.000000
(274, 5) (118, 5)
              count         mean         std  ...     50%      75%     max
cylinders     274.0     5.503650    1.720908  ...     4.0     8.00     8.0
displacement  274.0   196.131387  106.618440  ...   151.0   302.00   455.0
horsepower    274.0   104.755474   39.416747  ...    94.0   129.00   230.0
weight        274.0  2981.941606  863.904789  ...  2831.5  3641.75  4997.0

[4 rows x 8 columns]
222    17.0
247    39.4
Name: mpg, dtype: float64
1    15.0
2    18.0
Name: mpg, dtype: float64
Index(['cylinders', 'displacement', 'horsepower', 'weight'], dtype='object')

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense (Dense)               (None, 64)                320       
                                                                 
 dense_1 (Dense)             (None, 64)                4160      
                                                                 
 dense_2 (Dense)             (None, 1)                 65        
                                                                 
=================================================================
Total params: 4,545
Trainable params: 4,545
Non-trainable params: 0
_________________________________________________________________
None
Epoch 1/10000

1/7 [===>..........................] - ETA: 1s - loss: 680.4760 - mean_absolute_error: 24.7800 - mean_squared_error: 680.4760
7/7 [==============================] - 0s 21ms/step - loss: 601.4277 - mean_absolute_error: 23.1848 - mean_squared_error: 601.4277 - val_loss: 591.8640 - val_mean_absolute_error: 23.0248 - val_mean_squared_error: 591.8640
Epoch 2/10000

1/7 [===>..........................] - ETA: 0s - loss: 680.3566 - mean_absolute_error: 24.8924 - mean_squared_error: 680.3566
7/7 [==============================] - 0s 3ms/step - loss: 568.0007 - mean_absolute_error: 22.4028 - mean_squared_error: 568.0007 - val_loss: 560.7090 - val_mean_absolute_error: 22.3425 - val_mean_squared_error: 560.7090
Epoch 3/10000

1/7 [===>..........................] - ETA: 0s - loss: 419.8855 - mean_absolute_error: 19.3354 - mean_squared_error: 419.8855
7/7 [==============================] - 0s 3ms/step - loss: 534.2181 - mean_absolute_error: 21.6125 - mean_squared_error: 534.2181 - val_loss: 524.8778 - val_mean_absolute_error: 21.5590 - val_mean_squared_error: 524.8778
Epoch 4/10000

1/7 [===>..........................] - ETA: 0s - loss: 507.0336 - mean_absolute_error: 20.8416 - mean_squared_error: 507.0336
7/7 [==============================] - 0s 3ms/step - loss: 495.0857 - mean_absolute_error: 20.6901 - mean_squared_error: 495.0857 - val_loss: 483.7122 - val_mean_absolute_error: 20.6297 - val_mean_squared_error: 483.7122
Epoch 5/10000

1/7 [===>..........................] - ETA: 0s - loss: 408.2064 - mean_absolute_error: 18.7368 - mean_squared_error: 408.2064
7/7 [==============================] - 0s 3ms/step - loss: 450.4423 - mean_absolute_error: 19.5857 - mean_squared_error: 450.4423 - val_loss: 437.2163 - val_mean_absolute_error: 19.5276 - val_mean_squared_error: 437.2163
Epoch 6/10000

1/7 [===>..........................] - ETA: 0s - loss: 313.1874 - mean_absolute_error: 16.2679 - mean_squared_error: 313.1874
7/7 [==============================] - 0s 3ms/step - loss: 401.6044 - mean_absolute_error: 18.2996 - mean_squared_error: 401.6044 - val_loss: 386.7092 - val_mean_absolute_error: 18.2572 - val_mean_squared_error: 386.7092
Epoch 7/10000

1/7 [===>..........................] - ETA: 0s - loss: 319.1680 - mean_absolute_error: 15.8723 - mean_squared_error: 319.1680
7/7 [==============================] - 0s 3ms/step - loss: 350.2697 - mean_absolute_error: 16.8654 - mean_squared_error: 350.2697 - val_loss: 335.5231 - val_mean_absolute_error: 16.8724 - val_mean_squared_error: 335.5231
Epoch 8/10000

1/7 [===>..........................] - ETA: 0s - loss: 275.3731 - mean_absolute_error: 14.4855 - mean_squared_error: 275.3731
7/7 [==============================] - 0s 3ms/step - loss: 300.0649 - mean_absolute_error: 15.4424 - mean_squared_error: 300.0649 - val_loss: 285.6091 - val_mean_absolute_error: 15.4003 - val_mean_squared_error: 285.6091
Epoch 9/10000

1/7 [===>..........................] - ETA: 0s - loss: 159.2849 - mean_absolute_error: 10.4762 - mean_squared_error: 159.2849
7/7 [==============================] - 0s 3ms/step - loss: 251.0454 - mean_absolute_error: 14.0058 - mean_squared_error: 251.0454 - val_loss: 236.3739 - val_mean_absolute_error: 13.8193 - val_mean_squared_error: 236.3739
Epoch 10/10000

1/7 [===>..........................] - ETA: 0s - loss: 204.9026 - mean_absolute_error: 12.6882 - mean_squared_error: 204.9026
7/7 [==============================] - 0s 3ms/step - loss: 205.1691 - mean_absolute_error: 12.6148 - mean_squared_error: 205.1691 - val_loss: 191.8347 - val_mean_absolute_error: 12.3074 - val_mean_squared_error: 191.8347
Epoch 11/10000

1/7 [===>..........................] - ETA: 0s - loss: 172.3840 - mean_absolute_error: 10.6897 - mean_squared_error: 172.3840
7/7 [==============================] - 0s 3ms/step - loss: 164.0300 - mean_absolute_error: 11.1987 - mean_squared_error: 164.0300 - val_loss: 152.4515 - val_mean_absolute_error: 10.8554 - val_mean_squared_error: 152.4515
Epoch 12/10000

1/7 [===>..........................] - ETA: 0s - loss: 172.1983 - mean_absolute_error: 11.7179 - mean_squared_error: 172.1983
7/7 [==============================] - 0s 3ms/step - loss: 129.0225 - mean_absolute_error: 9.8662 - mean_squared_error: 129.0225 - val_loss: 121.2074 - val_mean_absolute_error: 9.5956 - val_mean_squared_error: 121.2074
Epoch 13/10000

1/7 [===>..........................] - ETA: 0s - loss: 126.9507 - mean_absolute_error: 9.9745 - mean_squared_error: 126.9507
7/7 [==============================] - 0s 3ms/step - loss: 99.7987 - mean_absolute_error: 8.4456 - mean_squared_error: 99.7987 - val_loss: 95.8570 - val_mean_absolute_error: 8.4000 - val_mean_squared_error: 95.8570
Epoch 14/10000

1/7 [===>..........................] - ETA: 0s - loss: 61.4363 - mean_absolute_error: 6.7040 - mean_squared_error: 61.4363
7/7 [==============================] - 0s 3ms/step - loss: 77.5760 - mean_absolute_error: 7.2028 - mean_squared_error: 77.5760 - val_loss: 77.8085 - val_mean_absolute_error: 7.3244 - val_mean_squared_error: 77.8085
Epoch 15/10000

1/7 [===>..........................] - ETA: 0s - loss: 78.8430 - mean_absolute_error: 6.5619 - mean_squared_error: 78.8430
7/7 [==============================] - 0s 3ms/step - loss: 62.9044 - mean_absolute_error: 6.2160 - mean_squared_error: 62.9044 - val_loss: 66.7750 - val_mean_absolute_error: 6.5801 - val_mean_squared_error: 66.7750
Epoch 16/10000

1/7 [===>..........................] - ETA: 0s - loss: 60.0901 - mean_absolute_error: 6.2861 - mean_squared_error: 60.0901
7/7 [==============================] - 0s 3ms/step - loss: 54.6862 - mean_absolute_error: 5.7036 - mean_squared_error: 54.6862 - val_loss: 61.4330 - val_mean_absolute_error: 6.2815 - val_mean_squared_error: 61.4330
Epoch 17/10000

1/7 [===>..........................] - ETA: 0s - loss: 60.5466 - mean_absolute_error: 6.7953 - mean_squared_error: 60.5466
7/7 [==============================] - 0s 3ms/step - loss: 50.6541 - mean_absolute_error: 5.4266 - mean_squared_error: 50.6541 - val_loss: 57.5465 - val_mean_absolute_error: 6.0370 - val_mean_squared_error: 57.5465
Epoch 18/10000

1/7 [===>..........................] - ETA: 0s - loss: 56.2741 - mean_absolute_error: 6.0906 - mean_squared_error: 56.2741
7/7 [==============================] - 0s 3ms/step - loss: 48.0193 - mean_absolute_error: 5.3197 - mean_squared_error: 48.0193 - val_loss: 54.9215 - val_mean_absolute_error: 5.8894 - val_mean_squared_error: 54.9215
Epoch 19/10000

1/7 [===>..........................] - ETA: 0s - loss: 51.2549 - mean_absolute_error: 5.8761 - mean_squared_error: 51.2549
7/7 [==============================] - 0s 3ms/step - loss: 45.9287 - mean_absolute_error: 5.2135 - mean_squared_error: 45.9287 - val_loss: 51.5483 - val_mean_absolute_error: 5.6825 - val_mean_squared_error: 51.5483
Epoch 20/10000

1/7 [===>..........................] - ETA: 0s - loss: 50.5600 - mean_absolute_error: 5.6437 - mean_squared_error: 50.5600
7/7 [==============================] - 0s 3ms/step - loss: 42.8736 - mean_absolute_error: 5.0280 - mean_squared_error: 42.8736 - val_loss: 48.2185 - val_mean_absolute_error: 5.4748 - val_mean_squared_error: 48.2185
Epoch 21/10000

1/7 [===>..........................] - ETA: 0s - loss: 34.4499 - mean_absolute_error: 4.3802 - mean_squared_error: 34.4499
7/7 [==============================] - 0s 3ms/step - loss: 40.3116 - mean_absolute_error: 4.9054 - mean_squared_error: 40.3116 - val_loss: 45.2562 - val_mean_absolute_error: 5.2730 - val_mean_squared_error: 45.2562
Epoch 22/10000

1/7 [===>..........................] - ETA: 0s - loss: 61.7068 - mean_absolute_error: 5.7318 - mean_squared_error: 61.7068
7/7 [==============================] - 0s 3ms/step - loss: 37.9759 - mean_absolute_error: 4.7480 - mean_squared_error: 37.9759 - val_loss: 41.9157 - val_mean_absolute_error: 5.0596 - val_mean_squared_error: 41.9157
Epoch 23/10000

1/7 [===>..........................] - ETA: 0s - loss: 31.6613 - mean_absolute_error: 4.6477 - mean_squared_error: 31.6613
7/7 [==============================] - 0s 3ms/step - loss: 35.2623 - mean_absolute_error: 4.6154 - mean_squared_error: 35.2623 - val_loss: 39.8304 - val_mean_absolute_error: 4.9478 - val_mean_squared_error: 39.8304
Epoch 24/10000

1/7 [===>..........................] - ETA: 0s - loss: 33.2401 - mean_absolute_error: 4.5983 - mean_squared_error: 33.2401
7/7 [==============================] - 0s 3ms/step - loss: 33.1273 - mean_absolute_error: 4.4238 - mean_squared_error: 33.1273 - val_loss: 36.5776 - val_mean_absolute_error: 4.7240 - val_mean_squared_error: 36.5776
Epoch 25/10000

1/7 [===>..........................] - ETA: 0s - loss: 31.0012 - mean_absolute_error: 4.6573 - mean_squared_error: 31.0012
7/7 [==============================] - 0s 3ms/step - loss: 30.7996 - mean_absolute_error: 4.3089 - mean_squared_error: 30.7996 - val_loss: 34.0954 - val_mean_absolute_error: 4.5112 - val_mean_squared_error: 34.0954
Epoch 26/10000

1/7 [===>..........................] - ETA: 0s - loss: 25.0821 - mean_absolute_error: 3.9926 - mean_squared_error: 25.0821
7/7 [==============================] - 0s 3ms/step - loss: 28.6433 - mean_absolute_error: 4.1386 - mean_squared_error: 28.6433 - val_loss: 31.3404 - val_mean_absolute_error: 4.2903 - val_mean_squared_error: 31.3404
Epoch 27/10000

1/7 [===>..........................] - ETA: 0s - loss: 19.4081 - mean_absolute_error: 3.5444 - mean_squared_error: 19.4081
7/7 [==============================] - 0s 3ms/step - loss: 26.9365 - mean_absolute_error: 4.0184 - mean_squared_error: 26.9365 - val_loss: 29.6591 - val_mean_absolute_error: 4.1595 - val_mean_squared_error: 29.6591
Epoch 28/10000

1/7 [===>..........................] - ETA: 0s - loss: 16.1090 - mean_absolute_error: 3.3709 - mean_squared_error: 16.1090
7/7 [==============================] - 0s 3ms/step - loss: 24.9158 - mean_absolute_error: 3.8511 - mean_squared_error: 24.9158 - val_loss: 27.4577 - val_mean_absolute_error: 3.9322 - val_mean_squared_error: 27.4577
Epoch 29/10000

1/7 [===>..........................] - ETA: 0s - loss: 34.0473 - mean_absolute_error: 4.1305 - mean_squared_error: 34.0473
7/7 [==============================] - 0s 3ms/step - loss: 23.2951 - mean_absolute_error: 3.6899 - mean_squared_error: 23.2951 - val_loss: 25.9998 - val_mean_absolute_error: 3.7651 - val_mean_squared_error: 25.9998
Epoch 30/10000

1/7 [===>..........................] - ETA: 0s - loss: 31.2338 - mean_absolute_error: 3.9500 - mean_squared_error: 31.2338
7/7 [==============================] - 0s 3ms/step - loss: 22.1446 - mean_absolute_error: 3.5575 - mean_squared_error: 22.1446 - val_loss: 24.6765 - val_mean_absolute_error: 3.6129 - val_mean_squared_error: 24.6765
Epoch 31/10000

1/7 [===>..........................] - ETA: 0s - loss: 15.6989 - mean_absolute_error: 3.0156 - mean_squared_error: 15.6989
7/7 [==============================] - 0s 3ms/step - loss: 20.6000 - mean_absolute_error: 3.4105 - mean_squared_error: 20.6000 - val_loss: 23.7181 - val_mean_absolute_error: 3.4574 - val_mean_squared_error: 23.7181
Epoch 32/10000

1/7 [===>..........................] - ETA: 0s - loss: 25.1083 - mean_absolute_error: 3.4224 - mean_squared_error: 25.1083
7/7 [==============================] - 0s 3ms/step - loss: 19.6939 - mean_absolute_error: 3.3010 - mean_squared_error: 19.6939 - val_loss: 22.7881 - val_mean_absolute_error: 3.3326 - val_mean_squared_error: 22.7881
Epoch 33/10000

1/7 [===>..........................] - ETA: 0s - loss: 14.9332 - mean_absolute_error: 3.1259 - mean_squared_error: 14.9332
7/7 [==============================] - 0s 2ms/step - loss: 18.9630 - mean_absolute_error: 3.1956 - mean_squared_error: 18.9630 - val_loss: 22.1572 - val_mean_absolute_error: 3.2348 - val_mean_squared_error: 22.1572
Epoch 34/10000

1/7 [===>..........................] - ETA: 0s - loss: 12.2150 - mean_absolute_error: 2.4114 - mean_squared_error: 12.2150
7/7 [==============================] - 0s 3ms/step - loss: 18.2997 - mean_absolute_error: 3.0920 - mean_squared_error: 18.2997 - val_loss: 21.2409 - val_mean_absolute_error: 3.1456 - val_mean_squared_error: 21.2409
Epoch 35/10000

1/7 [===>..........................] - ETA: 0s - loss: 11.4307 - mean_absolute_error: 2.7604 - mean_squared_error: 11.4307
7/7 [==============================] - 0s 3ms/step - loss: 17.8721 - mean_absolute_error: 3.1003 - mean_squared_error: 17.8721 - val_loss: 21.2036 - val_mean_absolute_error: 3.1207 - val_mean_squared_error: 21.2036
Epoch 36/10000

1/7 [===>..........................] - ETA: 0s - loss: 14.4703 - mean_absolute_error: 2.9405 - mean_squared_error: 14.4703
7/7 [==============================] - 0s 3ms/step - loss: 17.1524 - mean_absolute_error: 3.0119 - mean_squared_error: 17.1524 - val_loss: 21.4782 - val_mean_absolute_error: 3.1813 - val_mean_squared_error: 21.4782
Epoch 37/10000

1/7 [===>..........................] - ETA: 0s - loss: 13.6271 - mean_absolute_error: 2.9719 - mean_squared_error: 13.6271
7/7 [==============================] - 0s 3ms/step - loss: 17.1712 - mean_absolute_error: 2.9854 - mean_squared_error: 17.1712 - val_loss: 20.5797 - val_mean_absolute_error: 3.0848 - val_mean_squared_error: 20.5797
Epoch 38/10000

1/7 [===>..........................] - ETA: 0s - loss: 28.3013 - mean_absolute_error: 3.7912 - mean_squared_error: 28.3013
7/7 [==============================] - 0s 3ms/step - loss: 16.8312 - mean_absolute_error: 2.9674 - mean_squared_error: 16.8312 - val_loss: 20.3102 - val_mean_absolute_error: 3.0284 - val_mean_squared_error: 20.3102
Epoch 39/10000

1/7 [===>..........................] - ETA: 0s - loss: 14.7526 - mean_absolute_error: 3.1983 - mean_squared_error: 14.7526
7/7 [==============================] - 0s 3ms/step - loss: 16.8290 - mean_absolute_error: 2.9640 - mean_squared_error: 16.8290 - val_loss: 20.5047 - val_mean_absolute_error: 3.0471 - val_mean_squared_error: 20.5047
Epoch 40/10000

1/7 [===>..........................] - ETA: 0s - loss: 13.7946 - mean_absolute_error: 2.8730 - mean_squared_error: 13.7946
7/7 [==============================] - 0s 3ms/step - loss: 16.2755 - mean_absolute_error: 2.9209 - mean_squared_error: 16.2755 - val_loss: 20.2221 - val_mean_absolute_error: 3.0333 - val_mean_squared_error: 20.2221
Epoch 41/10000

1/7 [===>..........................] - ETA: 0s - loss: 15.1452 - mean_absolute_error: 2.7385 - mean_squared_error: 15.1452
7/7 [==============================] - 0s 3ms/step - loss: 15.8887 - mean_absolute_error: 2.8588 - mean_squared_error: 15.8887 - val_loss: 20.5933 - val_mean_absolute_error: 3.0518 - val_mean_squared_error: 20.5933
Epoch 42/10000

1/7 [===>..........................] - ETA: 0s - loss: 21.5418 - mean_absolute_error: 3.2976 - mean_squared_error: 21.5418
7/7 [==============================] - 0s 3ms/step - loss: 16.8122 - mean_absolute_error: 2.9391 - mean_squared_error: 16.8122 - val_loss: 19.9304 - val_mean_absolute_error: 2.9836 - val_mean_squared_error: 19.9304
Epoch 43/10000

1/7 [===>..........................] - ETA: 0s - loss: 13.5428 - mean_absolute_error: 2.8874 - mean_squared_error: 13.5428
7/7 [==============================] - 0s 3ms/step - loss: 15.9525 - mean_absolute_error: 2.8678 - mean_squared_error: 15.9525 - val_loss: 19.7977 - val_mean_absolute_error: 2.9853 - val_mean_squared_error: 19.7977
Epoch 44/10000

1/7 [===>..........................] - ETA: 0s - loss: 14.9289 - mean_absolute_error: 3.0643 - mean_squared_error: 14.9289
7/7 [==============================] - 0s 3ms/step - loss: 15.8257 - mean_absolute_error: 2.8693 - mean_squared_error: 15.8257 - val_loss: 20.1326 - val_mean_absolute_error: 3.0181 - val_mean_squared_error: 20.1326
Epoch 45/10000

1/7 [===>..........................] - ETA: 0s - loss: 9.5729 - mean_absolute_error: 2.5294 - mean_squared_error: 9.5729
7/7 [==============================] - 0s 3ms/step - loss: 16.0579 - mean_absolute_error: 2.8986 - mean_squared_error: 16.0579 - val_loss: 19.8492 - val_mean_absolute_error: 2.9720 - val_mean_squared_error: 19.8492
Epoch 46/10000

1/7 [===>..........................] - ETA: 0s - loss: 19.3899 - mean_absolute_error: 3.0403 - mean_squared_error: 19.3899
7/7 [==============================] - 0s 3ms/step - loss: 15.9067 - mean_absolute_error: 2.8749 - mean_squared_error: 15.9067 - val_loss: 19.9324 - val_mean_absolute_error: 2.9916 - val_mean_squared_error: 19.9324
Epoch 47/10000

1/7 [===>..........................] - ETA: 0s - loss: 13.0213 - mean_absolute_error: 2.6371 - mean_squared_error: 13.0213
7/7 [==============================] - 0s 3ms/step - loss: 15.7572 - mean_absolute_error: 2.8664 - mean_squared_error: 15.7572 - val_loss: 20.0052 - val_mean_absolute_error: 3.0032 - val_mean_squared_error: 20.0052
Epoch 48/10000

1/7 [===>..........................] - ETA: 0s - loss: 11.9728 - mean_absolute_error: 2.6327 - mean_squared_error: 11.9728
7/7 [==============================] - 0s 3ms/step - loss: 15.6571 - mean_absolute_error: 2.8415 - mean_squared_error: 15.6571 - val_loss: 19.6889 - val_mean_absolute_error: 2.9865 - val_mean_squared_error: 19.6889
Epoch 49/10000

1/7 [===>..........................] - ETA: 0s - loss: 22.8097 - mean_absolute_error: 3.4804 - mean_squared_error: 22.8097
7/7 [==============================] - 0s 3ms/step - loss: 15.5446 - mean_absolute_error: 2.8439 - mean_squared_error: 15.5446 - val_loss: 19.6736 - val_mean_absolute_error: 2.9819 - val_mean_squared_error: 19.6736
Epoch 50/10000

1/7 [===>..........................] - ETA: 0s - loss: 23.2874 - mean_absolute_error: 3.5320 - mean_squared_error: 23.2874
7/7 [==============================] - 0s 3ms/step - loss: 15.6644 - mean_absolute_error: 2.8380 - mean_squared_error: 15.6644 - val_loss: 19.5981 - val_mean_absolute_error: 2.9795 - val_mean_squared_error: 19.5981
Epoch 51/10000

1/7 [===>..........................] - ETA: 0s - loss: 12.2018 - mean_absolute_error: 2.8134 - mean_squared_error: 12.2018
7/7 [==============================] - 0s 3ms/step - loss: 16.1080 - mean_absolute_error: 2.8916 - mean_squared_error: 16.1080 - val_loss: 19.6137 - val_mean_absolute_error: 2.9753 - val_mean_squared_error: 19.6137
Epoch 52/10000

1/7 [===>..........................] - ETA: 0s - loss: 19.6972 - mean_absolute_error: 3.0819 - mean_squared_error: 19.6972
7/7 [==============================] - 0s 3ms/step - loss: 15.2194 - mean_absolute_error: 2.8007 - mean_squared_error: 15.2194 - val_loss: 19.6220 - val_mean_absolute_error: 2.9872 - val_mean_squared_error: 19.6220
Epoch 53/10000

1/7 [===>..........................] - ETA: 0s - loss: 15.3694 - mean_absolute_error: 2.7427 - mean_squared_error: 15.3694
7/7 [==============================] - 0s 3ms/step - loss: 15.3328 - mean_absolute_error: 2.8302 - mean_squared_error: 15.3328 - val_loss: 19.5572 - val_mean_absolute_error: 2.9706 - val_mean_squared_error: 19.5572
Epoch 54/10000

1/7 [===>..........................] - ETA: 0s - loss: 14.3996 - mean_absolute_error: 2.8174 - mean_squared_error: 14.3996
7/7 [==============================] - 0s 3ms/step - loss: 15.6872 - mean_absolute_error: 2.8493 - mean_squared_error: 15.6872 - val_loss: 19.5703 - val_mean_absolute_error: 2.9711 - val_mean_squared_error: 19.5703
Epoch 55/10000

1/7 [===>..........................] - ETA: 0s - loss: 16.6972 - mean_absolute_error: 3.0000 - mean_squared_error: 16.6972
7/7 [==============================] - 0s 2ms/step - loss: 15.3090 - mean_absolute_error: 2.8341 - mean_squared_error: 15.3090 - val_loss: 19.5212 - val_mean_absolute_error: 2.9674 - val_mean_squared_error: 19.5212
Epoch 56/10000

1/7 [===>..........................] - ETA: 0s - loss: 15.5535 - mean_absolute_error: 2.8565 - mean_squared_error: 15.5535
7/7 [==============================] - 0s 3ms/step - loss: 15.0543 - mean_absolute_error: 2.8025 - mean_squared_error: 15.0543 - val_loss: 19.4893 - val_mean_absolute_error: 2.9645 - val_mean_squared_error: 19.4893
Epoch 57/10000

1/7 [===>..........................] - ETA: 0s - loss: 12.4237 - mean_absolute_error: 2.7484 - mean_squared_error: 12.4237
7/7 [==============================] - 0s 3ms/step - loss: 15.0483 - mean_absolute_error: 2.8079 - mean_squared_error: 15.0483 - val_loss: 20.2153 - val_mean_absolute_error: 3.0010 - val_mean_squared_error: 20.2153
Epoch 58/10000

1/7 [===>..........................] - ETA: 0s - loss: 13.5162 - mean_absolute_error: 2.6402 - mean_squared_error: 13.5162
7/7 [==============================] - 0s 3ms/step - loss: 15.3872 - mean_absolute_error: 2.8026 - mean_squared_error: 15.3872 - val_loss: 19.6546 - val_mean_absolute_error: 2.9706 - val_mean_squared_error: 19.6546
Epoch 59/10000

1/7 [===>..........................] - ETA: 0s - loss: 15.3013 - mean_absolute_error: 2.8112 - mean_squared_error: 15.3013
7/7 [==============================] - 0s 3ms/step - loss: 15.3402 - mean_absolute_error: 2.7953 - mean_squared_error: 15.3402 - val_loss: 19.5386 - val_mean_absolute_error: 2.9649 - val_mean_squared_error: 19.5386
Epoch 60/10000

1/7 [===>..........................] - ETA: 0s - loss: 7.6489 - mean_absolute_error: 1.9015 - mean_squared_error: 7.6489
7/7 [==============================] - 0s 3ms/step - loss: 15.0646 - mean_absolute_error: 2.8211 - mean_squared_error: 15.0646 - val_loss: 19.6878 - val_mean_absolute_error: 2.9713 - val_mean_squared_error: 19.6878
Epoch 61/10000

1/7 [===>..........................] - ETA: 0s - loss: 13.3874 - mean_absolute_error: 2.9319 - mean_squared_error: 13.3874
7/7 [==============================] - 0s 3ms/step - loss: 15.0976 - mean_absolute_error: 2.8221 - mean_squared_error: 15.0976 - val_loss: 19.9646 - val_mean_absolute_error: 2.9790 - val_mean_squared_error: 19.9646
          loss  ...  val_mean_squared_error
0   601.427673  ...              591.864014
1   568.000671  ...              560.709045
2   534.218140  ...              524.877808
3   495.085724  ...              483.712219
4   450.442322  ...              437.216309
..         ...  ...                     ...
56   15.048346  ...               20.215349
57   15.387167  ...               19.654648
58   15.340160  ...               19.538551
59   15.064593  ...               19.687771
60   15.097616  ...               19.964642

[61 rows x 6 columns]

1/4 [======>.......................] - ETA: 0s - loss: 12.4303 - mean_absolute_error: 2.5080 - mean_squared_error: 12.4303
4/4 [==============================] - 0s 1ms/step - loss: 14.7585 - mean_absolute_error: 3.0017 - mean_squared_error: 14.7585
loss : 14.758
mae : 3.002
mse : 14.758

1/4 [======>.......................] - ETA: 0s
4/4 [==============================] - 0s 835us/step
설명력 :  0.7201497337303957

EarlyStopping() 안에 patience의 의미는 val_loss가 n번 이내에 떨어지지 않으면 조기 종료 하겠다는 의미이다.

 

train / test split을 함수를 사용하지 않고 직접 수식으로 사용하였다. 표준화 또한 수식을 사용하였다.

모델 설계를 함수로 만들었다. 재사용하기 편하다.

fit() 전에 모델을 실행해도 됨. 다만 성능은 기대하지 않는다.

 

 

상관계수 시각화

 

mse, mae 시각화

조기 종료는 오버피팅을 방지할 수 있다.