1
2
3
4
5
6
7
8

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

1
2
3
4
5
6
7
8
9
10
11
12
13

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


column_names = ['MPG','气缸','排量','马力','重量','加速度', '年份', '产地']
raw_dataset = pd.read_csv(dataset_path, names=column_names,
                      na_values = "?", comment='\t',
                      sep=" ", skipinitialspace=True)

dataset = raw_dataset.copy()

dataset.head(3)

1

dataset.isna().sum()

1
2
3
4
5
6
7
8
9

MPG    0
气缸     0
排量     0
马力     6
重量     0
加速度    0
年份     0
产地    0
dtype: int64

1

dataset = dataset.dropna()

1
2
3
4
5
6

origin = dataset.pop('产地')
dataset['美国'] = (origin == 1)*1.0
dataset['欧洲'] = (origin == 2)*1.0
dataset['日本'] = (origin == 3)*1.0

dataset.head(3)

1
2
3

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

1
2
3
4
5

plt.rcParams['font.sans-serif']=['SimHei']
plt.rcParams['axes.unicode_minus']=False

sns.pairplot(train_dataset[["MPG", "气缸", "排量", "重量"]], diag_kind="kde")

1
2
3
4

train_stats = train_dataset.describe()
train_stats.pop("MPG")
train_stats = train_stats.transpose()
train_stats

1
2
3

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

1
2
3
4

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

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

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

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

model = build_model()

model.summary()

1
2
3
4
5
6
7
8
9
10
11
12
13
14

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param 
=================================================================
dense_4 (Dense)              (None, 64)                640       
_________________________________________________________________
dense_5 (Dense)              (None, 32)                2080      
_________________________________________________________________
dense_6 (Dense)              (None, 1)                 33        
=================================================================
Total params: 2,753
Trainable params: 2,753
Non-trainable params: 0
_________________________________________________________________

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

import sys

    
EPOCHS = 1000

class ProgressBar(keras.callbacks.Callback):
  def on_epoch_end(self, epoch, logs):
    
    self.draw_progress_bar(epoch + 1, EPOCHS)

  def draw_progress_bar(self, cur, total, bar_len=50):
    cur_len = int(cur / total * bar_len)
    sys.stdout.write("\r")
    sys.stdout.write("[{:<{}}] {}/{}".format("=" * cur_len, bar_len, cur, total))
    sys.stdout.flush()

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

1

[==================================================] 1000/1000

1
2
3

hist = pd.DataFrame(history.history)
hist['epoch'] = history.epoch
hist.tail(3)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

def plot_history(history):
  hist = pd.DataFrame(history.history)
  hist['epoch'] = history.epoch
  plt.figure()
  plt.xlabel('epoch')
  plt.ylabel('metric - MSE')
  plt.plot(hist['epoch'], hist['mse'], label='训练集')
  plt.plot(hist['epoch'], hist['val_mse'], label = '验证集')
  plt.ylim([0, 20])
  plt.legend()
  
  plt.figure()
  plt.xlabel('epoch')
  plt.ylabel('metric - MAE')
  plt.plot(hist['epoch'], hist['mae'], label='训练集')
  plt.plot(hist['epoch'], hist['val_mae'], label = '验证集')
  plt.ylim([0, 5])
  plt.legend()
  
plot_history(history)

1
2
3
4
5
6

model = build_model()
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, ProgressBar()])
plot_history(history)

1

[===                                               ] 70/1000

1
2
3

loss, mae, mse = model.evaluate(normed_test_data, test_labels, verbose=0)
print("测试集平均绝对误差(MAE): {:5.2f} MPG".format(mae))

1
2
3
4
5
6
7
8
9
10

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

plt.scatter(test_labels, test_pred)
plt.xlabel('真实值')
plt.ylabel('预测值')
plt.axis('equal')
plt.axis('square')
plt.xlim([0,plt.xlim()[1]])
plt.ylim([0,plt.ylim()[1]])
plt.plot([-100, 100], [-100, 100])