Code_자동차 연비예측_케라스편

케라스(Keras) 코드

앞서 요약했던 DNN과 관련하여 모델을 만들고 예측하는 코드다.

데이터는 자동차 연비와 관련되어 있으며 이를 예측하는 모델과 실제값을 비교하는 내용이다.

import pathlib

import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers

print(tf.__version__)

2.3.0

데이터셋

• 1970년대 후반과 1980년대 초반의 자동차 연비를 예측하는 모델
• 다운받고 열 지정

dataset_path = keras.utils.get_file("auto-mpg.data", "http://archive.ics.uci.edu/ml/machine-learning-databases/auto-mpg/auto-mpg.data")
dataset_path

'C:\\Users\\uos\\.keras\\datasets\\auto-mpg.data'

column_names = ['MPG','Cylinders','Displacement','Horsepower','Weight',
                'Acceleration', 'Model Year', 'Origin']
raw_dataset = pd.read_csv(dataset_path, names=column_names,
                      na_values = "?", comment='\t',
                      sep=" ", skipinitialspace=True)

dataset = raw_dataset.copy()
dataset.tail()

	MPG	Cylinders	Displacement	Horsepower	Weight	Acceleration	Model Year	Origin
393	27.0	4	140.0	86.0	2790.0	15.6	82	1
394	44.0	4	97.0	52.0	2130.0	24.6	82	2
395	32.0	4	135.0	84.0	2295.0	11.6	82	1
396	28.0	4	120.0	79.0	2625.0	18.6	82	1
397	31.0	4	119.0	82.0	2720.0	19.4	82	1

데이터 정제

dataset.isna().sum()

MPG             0
Cylinders       0
Displacement    0
Horsepower      6
Weight          0
Acceleration    0
Model Year      0
Origin          0
dtype: int64

#결측치 데이터
dataset[dataset['Horsepower'].isnull()]

	MPG	Cylinders	Displacement	Horsepower	Weight	Acceleration	Model Year	Origin
32	25.0	4	98.0	NaN	2046.0	19.0	71	1
126	21.0	6	200.0	NaN	2875.0	17.0	74	1
330	40.9	4	85.0	NaN	1835.0	17.3	80	2
336	23.6	4	140.0	NaN	2905.0	14.3	80	1
354	34.5	4	100.0	NaN	2320.0	15.8	81	2
374	23.0	4	151.0	NaN	3035.0	20.5	82	1

#결측치 평균으로 대체
dataset=dataset.fillna(dataset['Horsepower'].mean())
dataset[330:337]

	MPG	Cylinders	Displacement	Horsepower	Weight	Acceleration	Model Year	Origin
330	40.9	4	85.0	104.469388	1835.0	17.3	80	2
331	33.8	4	97.0	67.000000	2145.0	18.0	80	3
332	29.8	4	89.0	62.000000	1845.0	15.3	80	2
333	32.7	6	168.0	132.000000	2910.0	11.4	80	3
334	23.7	3	70.0	100.000000	2420.0	12.5	80	3
335	35.0	4	122.0	88.000000	2500.0	15.1	80	2
336	23.6	4	140.0	104.469388	2905.0	14.3	80	1

원핫인코딩

• origin (1,2,3) : 범주형

dataset.groupby('Origin').count()

	MPG	Cylinders	Displacement	Horsepower	Weight	Acceleration	Model Year
Origin
1	249	249	249	249	249	249	249
2	70	70	70	70	70	70	70
3	79	79	79	79	79	79	79

origin = dataset.pop('Origin')  # 레이블로 받음

dataset['USA'] = (origin == 1)*1.0
dataset['Europe'] = (origin == 2)*1.0
dataset['Japan'] = (origin == 3)*1.0
dataset.tail()

	MPG	Cylinders	Displacement	Horsepower	Weight	Acceleration	Model Year	USA	Europe	Japan
393	27.0	4	140.0	86.0	2790.0	15.6	82	1.0	0.0	0.0
394	44.0	4	97.0	52.0	2130.0	24.6	82	0.0	1.0	0.0
395	32.0	4	135.0	84.0	2295.0	11.6	82	1.0	0.0	0.0
396	28.0	4	120.0	79.0	2625.0	18.6	82	1.0	0.0	0.0
397	31.0	4	119.0	82.0	2720.0	19.4	82	1.0	0.0	0.0

데이터 셋 분할

train_dataset = dataset.sample(frac=0.8,random_state=0)
test_dataset = dataset.drop(train_dataset.index)

시각화

sns.pairplot(train_dataset[["MPG", "Cylinders", "Displacement", "Weight"]], diag_kind="kde")

<seaborn.axisgrid.PairGrid at 0x18a81e1a490>

train_stats = train_dataset.describe()
train_stats

	MPG	Cylinders	Displacement	Horsepower	Weight	Acceleration	Model Year	USA	Europe	Japan
count	318.000000	318.000000	318.000000	318.000000	318.000000	318.000000	318.000000	318.000000	318.000000	318.000000
mean	23.590566	5.427673	193.061321	104.073418	2963.823899	15.595912	75.946541	0.641509	0.163522	0.194969
std	7.913617	1.682941	103.812742	38.368477	844.749805	2.796282	3.705266	0.480313	0.370424	0.396801
min	10.000000	3.000000	70.000000	46.000000	1613.000000	8.000000	70.000000	0.000000	0.000000	0.000000
25%	17.125000	4.000000	100.250000	75.250000	2219.250000	13.900000	73.000000	0.000000	0.000000	0.000000
50%	22.750000	4.000000	151.000000	92.000000	2792.500000	15.500000	76.000000	1.000000	0.000000	0.000000
75%	29.000000	6.000000	259.500000	120.000000	3571.250000	17.300000	79.000000	1.000000	0.000000	0.000000
max	46.600000	8.000000	455.000000	230.000000	5140.000000	24.800000	82.000000	1.000000	1.000000	1.000000

train_stats.pop("MPG")  # mpg(자동차 연비: y)
train_stats = train_stats.transpose()
train_stats  # 값의 범위가 크기때문에 정규화 진행 필요

	count	mean	std	min	25%	50%	75%	max
Cylinders	318.0	5.427673	1.682941	3.0	4.00	4.0	6.00	8.0
Displacement	318.0	193.061321	103.812742	70.0	100.25	151.0	259.50	455.0
Horsepower	318.0	104.073418	38.368477	46.0	75.25	92.0	120.00	230.0
Weight	318.0	2963.823899	844.749805	1613.0	2219.25	2792.5	3571.25	5140.0
Acceleration	318.0	15.595912	2.796282	8.0	13.90	15.5	17.30	24.8
Model Year	318.0	75.946541	3.705266	70.0	73.00	76.0	79.00	82.0
USA	318.0	0.641509	0.480313	0.0	0.00	1.0	1.00	1.0
Europe	318.0	0.163522	0.370424	0.0	0.00	0.0	0.00	1.0
Japan	318.0	0.194969	0.396801	0.0	0.00	0.0	0.00	1.0

특성과 레이블 분리

train_labels = train_dataset.pop('MPG')
test_labels = test_dataset.pop('MPG')

정규화

def norm(x):
  return (x - train_stats['mean']) / train_stats['std']
normed_train_data = norm(train_dataset)
normed_test_data = norm(test_dataset)

모델 만들기

• densely connected 2개의 층

def build_model():
    model =keras.Sequential([
        layers.Dense(64,activation='relu',input_shape=[len(train_dataset.keys())]),
        layers.Dense(64, activation='relu'),
        layers.Dense(1)
    ])

    optimizer=tf.keras.optimizers.Adam(0.001) 
    
    model.compile(loss='mse', optimizer=optimizer,metrics=['mae','mse']) # 분류일때,'accuracy'
    
    return model

model=build_model()

model.summary()

Model: "sequential_2"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_6 (Dense)              (None, 64)                640       
_________________________________________________________________
dense_7 (Dense)              (None, 64)                4160      
_________________________________________________________________
dense_8 (Dense)              (None, 1)                 65        
=================================================================
Total params: 4,865
Trainable params: 4,865
Non-trainable params: 0
_________________________________________________________________

example_batch = normed_train_data[:10]
example_result = model.predict(example_batch)
example_result

array([[ 0.0416676 ],
       [-0.03854847],
       [-0.11580162],
       [-0.17297746],
       [ 0.00868107],
       [-0.3099752 ],
       [-0.3315743 ],
       [ 0.02203141],
       [ 0.15218496],
       [ 0.0900453 ]], dtype=float32)

모델 훈련

• 1000번의 epoch동안 훈련

# 에포크가 끝날 때마다 점(.)을 출력해 훈련 진행 과정을 표시합니다
class PrintDot(keras.callbacks.Callback):
  def on_epoch_end(self, epoch, logs):
    if epoch % 100 == 0: print('')
    print('.', end='')

EPOCHS = 1000

history = model.fit(
  normed_train_data, train_labels,
  epochs=EPOCHS, validation_split = 0.2, verbose=0,
  callbacks=[PrintDot()])

....................................................................................................
....................................................................................................
....................................................................................................
....................................................................................................
....................................................................................................
....................................................................................................
....................................................................................................
....................................................................................................
....................................................................................................
....................................................................................................

hist = pd.DataFrame(history.history)
hist['epoch'] = history.epoch
hist.tail()

	loss	mae	mse	val_loss	val_mae	val_mse	epoch
995	2.358095	1.010328	2.358095	9.766920	2.035713	9.766920	995
996	2.424612	1.034429	2.424612	9.920653	2.076196	9.920653	996
997	2.260907	1.004899	2.260907	9.677170	2.016382	9.677170	997
998	2.379298	1.019345	2.379298	9.484661	2.004562	9.484661	998
999	2.322387	0.990800	2.322387	9.923290	2.072285	9.923290	999

import matplotlib.pyplot as plt

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['mae'],
           label='Train Error')
  plt.plot(hist['epoch'], hist['val_mae'],
           label = 'Val Error')
  plt.ylim([0,5])
  plt.legend()

  plt.subplot(2,1,2)
  plt.xlabel('Epoch')
  plt.ylabel('Mean Square Error [$MPG^2$]')
  plt.plot(hist['epoch'], hist['mse'],
           label='Train Error')
  plt.plot(hist['epoch'], hist['val_mse'],
           label = 'Val Error')
  plt.ylim([0,20])
  plt.legend()
  plt.show()

plot_history(history)

train과 val error의 차이가 많이 나서 지정된 에포크 횟수동안 성능향상이 없으면 자동으로 훈련을 멈춤

model = build_model()

# patience 매개변수는 성능 향상을 체크할 에포크 횟수입니다
early_stop = keras.callbacks.EarlyStopping(monitor='val_loss', patience=10)

history = model.fit(normed_train_data, train_labels, epochs=EPOCHS,
                    validation_split = 0.2, verbose=0, callbacks=[early_stop, PrintDot()])

plot_history(history)

....................................................................................................
...........................................................

테스트세트로 평가

loss, mae, mse = model.evaluate(normed_test_data, test_labels, verbose=2)

print("테스트 세트의 평균 절대 오차: {:5.2f} MPG".format(mae))

3/3 - 0s - loss: 6.5907 - mae: 1.9264 - mse: 6.5907
테스트 세트의 평균 절대 오차:  1.93 MPG

예측

• 테스트 세트(실제 y)에서 MPG값 예측 비교

test_predictions = model.predict(normed_test_data).flatten()

plt.scatter(test_labels, test_predictions)
plt.xlabel('True Values [MPG]')
plt.ylabel('Predictions [MPG]')
plt.axis('equal')
plt.axis('square')
plt.xlim([0,plt.xlim()[1]])
plt.ylim([0,plt.ylim()[1]])
_ = plt.plot([-100, 100], [-100, 100])

error = test_predictions - test_labels
plt.hist(error, bins = 25)
plt.xlabel("Prediction Error [MPG]")
_ = plt.ylabel("Count")

참고: TensorFlow 자동차 연비 예측하기 : 회귀