TensorFlow 从入门到精通（2）—— 波士顿房价预测

import tensorflow as tf

• 防止过拟合的一个办法是设置验证集，当模型在训练集上训练完之后，利用验证集对模型进行调参优化，最终用测试集测试模型性能

print(tf.__version__)

2.5.1

一、准备数据集

import pandas as pd

df = pd.read_csv('boston.csv')
df.head()

	CRIM	ZN	INDUS	CHAS	NOX	RM	AGE	DIS	RAD	TAX	PTRATIO	LSTAT	MEDV
0	0.00632	18.0	2.31	0	0.538	6.575	65.2	4.0900	1	296	15.3	4.98	24.0
1	0.02731	0.0	7.07	0	0.469	6.421	78.9	4.9671	2	242	17.8	9.14	21.6
2	0.02729	0.0	7.07	0	0.469	7.185	61.1	4.9671	2	242	17.8	4.03	34.7
3	0.03237	0.0	2.18	0	0.458	6.998	45.8	6.0622	3	222	18.7	2.94	33.4
4	0.06905	0.0	2.18	0	0.458	7.147	54.2	6.0622	3	222	18.7	5.33	36.2

df.tail()

	CRIM	ZN	INDUS	CHAS	NOX	RM	AGE	DIS	RAD	TAX	PTRATIO	LSTAT	MEDV
501	0.06263	0.0	11.93	0	0.573	6.593	69.1	2.4786	1	273	21.0	9.67	22.4
502	0.04527	0.0	11.93	0	0.573	6.120	76.7	2.2875	1	273	21.0	9.08	20.6
503	0.06076	0.0	11.93	0	0.573	6.976	91.0	2.1675	1	273	21.0	5.64	23.9
504	0.10959	0.0	11.93	0	0.573	6.794	89.3	2.3889	1	273	21.0	6.48	22.0
505	0.04741	0.0	11.93	0	0.573	6.030	80.8	2.5050	1	273	21.0	7.88	11.9

df.describe()

	CRIM	ZN	INDUS	CHAS	NOX	RM	AGE	DIS	RAD	TAX	PTRATIO	LSTAT	MEDV
count	506.000000	506.000000	506.000000	506.000000	506.000000	506.000000	506.000000	506.000000	506.000000	506.000000	506.000000	506.000000	506.000000
mean	3.613524	11.363636	11.136779	0.069170	0.554695	6.284634	68.574901	3.795043	9.549407	408.237154	18.455534	12.653063	22.532806
std	8.601545	23.322453	6.860353	0.253994	0.115878	0.702617	28.148861	2.105710	8.707259	168.537116	2.164946	7.141062	9.197104
min	0.006320	0.000000	0.460000	0.000000	0.385000	3.561000	2.900000	1.129600	1.000000	187.000000	12.600000	1.730000	5.000000
25%	0.082045	0.000000	5.190000	0.000000	0.449000	5.885500	45.025000	2.100175	4.000000	279.000000	17.400000	6.950000	17.025000
50%	0.256510	0.000000	9.690000	0.000000	0.538000	6.208500	77.500000	3.207450	5.000000	330.000000	19.050000	11.360000	21.200000
75%	3.677082	12.500000	18.100000	0.000000	0.624000	6.623500	94.075000	5.188425	24.000000	666.000000	20.200000	16.955000	25.000000
max	88.976200	100.000000	27.740000	1.000000	0.871000	8.780000	100.000000	12.126500	24.000000	711.000000	22.000000	37.970000	50.000000

data = df.values
data.shape

(506, 13)

x_data = data[:,:12]
y_data = data[:,12]

from sklearn.preprocessing import scale
# 划分训练集，测试集，验证集
train_num = 300 # 训练集
valid_num = 100 # 验证集

# 训练集
x_train = x_data[:train_num]
y_train = y_data[:train_num]

# 验证集
x_valid = x_data[train_num:train_num+valid_num]
y_valid = y_data[train_num:train_num+valid_num]

# 测试集
x_test = x_data[train_num+valid_num:]
y_test = y_data[train_num+valid_num:]


# 转换数据类型 
'''
tf.cast可以直接将numpy数组转换为tensor
'''
x_train = tf.cast(scale(x_train),dtype=tf.float32)
x_valid = tf.cast(scale(x_valid),dtype=tf.float32)
x_test = tf.cast(scale(x_test),dtype=tf.float32)

二、模型

import numpy as np
def line_model(x,w,b):
    return tf.matmul(x,w) + b

三、训练

W = tf.Variable(tf.random.normal(shape=(12,1)))
B = tf.Variable(tf.zeros(1),dtype=tf.float32)

def loss(x,y,w,b):
    loss_ = tf.square(line_model(x,w,b) - y)
    return tf.reduce_mean(loss_)
    
def grand(x,y,w,b):
    with tf.GradientTape() as tape:
        loss_ = loss(x,y,w,b)
    return tape.gradient(loss_,[w,b])

training_epochs = 50
learning_rate = 0.001
batch_size = 10 # 训练一次的样本数
optimizer = tf.keras.optimizers.SGD(learning_rate)

total_steps = train_num // batch_size
loss_train_list = []
loss_valid_list = []
for epoch in range(training_epochs):
    for step in range(total_steps):
        xs = x_train[step*batch_size:(step+1)*batch_size]
        ys = y_train[step*batch_size:(step+1)*batch_size]
        
        grads = grand(xs,ys,W,B)
        optimizer.apply_gradients(zip(grads,[W,B]))
    loss_train = loss(x_train,y_train,W,B).numpy()
    loss_valid = loss(x_valid,y_valid,W,B).numpy()
    loss_train_list.append(loss_train)
    loss_valid_list.append(loss_valid)
    print(f'{
       epoch+1}次 loss_train:{
       loss_train} loss_valid:{
       loss_valid}')

1次 loss_train:665.7977905273438 loss_valid:477.3428039550781
2次 loss_train:593.6646728515625 loss_valid:413.1649475097656
3次 loss_train:533.6004028320312 loss_valid:362.72601318359375
4次 loss_train:482.3865661621094 loss_valid:321.768310546875
5次 loss_train:438.0271911621094 loss_valid:287.7205810546875
6次 loss_train:399.220703125 loss_valid:258.9722900390625
7次 loss_train:365.0664367675781 loss_valid:234.4657745361328
8次 loss_train:334.9011535644531 loss_valid:213.4668426513672
9次 loss_train:308.20770263671875 loss_valid:195.4357147216797
10次 loss_train:284.56365966796875 loss_valid:179.9535675048828
11次 loss_train:263.6124572753906 loss_valid:166.6811065673828
12次 loss_train:245.04588317871094 loss_valid:155.3340301513672
13次 loss_train:228.59408569335938 loss_valid:145.66885375976562
14次 loss_train:214.01869201660156 loss_valid:137.47401428222656
15次 loss_train:201.10826110839844 loss_valid:130.56398010253906
16次 loss_train:189.67495727539062 loss_valid:124.77549743652344
17次 loss_train:179.55177307128906 loss_valid:119.96448516845703
18次 loss_train:170.58995056152344 loss_valid:116.00358581542969
19次 loss_train:162.65756225585938 loss_valid:112.78060913085938
20次 loss_train:155.6372833251953 loss_valid:110.1964340209961
21次 loss_train:149.42471313476562 loss_valid:108.16374969482422
22次 loss_train:143.92759704589844 loss_valid:106.60588836669922
23次 loss_train:139.0637664794922 loss_valid:105.45535278320312
24次 loss_train:134.7605743408203 loss_valid:104.65306091308594
25次 loss_train:130.95376586914062 loss_valid:104.14726257324219
26次 loss_train:127.58616638183594 loss_valid:103.89268493652344
27次 loss_train:124.60723114013672 loss_valid:103.84986114501953
28次 loss_train:121.97221374511719 loss_valid:103.98436737060547
29次 loss_train:119.64164733886719 loss_valid:104.26632690429688
30次 loss_train:117.58037567138672 loss_valid:104.66973876953125
31次 loss_train:115.75743103027344 loss_valid:105.17220306396484
32次 loss_train:114.14533233642578 loss_valid:105.7542495727539
33次 loss_train:112.71986389160156 loss_valid:106.39917755126953
34次 loss_train:111.45946502685547 loss_valid:107.09259796142578
35次 loss_train:110.34514617919922 loss_valid:107.82219696044922
36次 loss_train:109.360107421875 loss_valid:108.57747650146484
37次 loss_train:108.4894790649414 loss_valid:109.3494873046875
38次 loss_train:107.72004699707031 loss_valid:110.13068389892578
39次 loss_train:107.04024505615234 loss_valid:110.91461181640625
40次 loss_train:106.43974304199219 loss_valid:111.69599914550781
41次 loss_train:105.909423828125 loss_valid:112.47042846679688
42次 loss_train:105.44117736816406 loss_valid:113.23422241210938
43次 loss_train:105.0279312133789 loss_valid:113.9843978881836
44次 loss_train:104.66329193115234 loss_valid:114.7186508178711
45次 loss_train:104.34174346923828 loss_valid:115.43498992919922
46次 loss_train:104.05831146240234 loss_valid:116.13207244873047
47次 loss_train:103.80862426757812 loss_valid:116.80872344970703
48次 loss_train:103.58882141113281 loss_valid:117.46420288085938
49次 loss_train:103.39546966552734 loss_valid:118.0979995727539
50次 loss_train:103.22554779052734 loss_valid:118.70991516113281

# 可视化损失
import matplotlib.pyplot as plt
%matplotlib inline
plt.plot(loss_train_list)
plt.plot(loss_valid_list)

[]

四、预测

W.numpy()

array([[ 0.04078581],
       [ 0.42111152],
       [-1.266555  ],
       [ 0.7909989 ],
       [-0.7962972 ],
       [ 1.6725844 ],
       [-0.23721075],
       [-0.900601  ],
       [ 0.45604745],
       [-0.51313245],
       [-2.141279  ],
       [-0.713085  ]], dtype=float32)

B.numpy()

array([24.265219], dtype=float32)

# 从测试集里随机挑一个
house_id = np.random.randint(0,x_data.shape[0]-train_num-valid_num)
# 真实值
y = y_test[house_id]
# 预测值
y_pre = int(line_model(x_test,W.numpy(),B.numpy())[house_id])

print(f'真实值：{
       y}，预测值：{
       y_pre}')

真实值：10.2，预测值：24

plt.plot(y_test,'r')
plt.plot(line_model(x_test,W.numpy(),B.numpy()),'g')

[]

plt.plot(y_train,'r')
plt.plot(line_model(x_train,W.numpy(),B.numpy()),'g')

[]

plt.plot(y_valid,'r')
plt.plot(line_model(x_valid,W.numpy(),B.numpy()),'g')

[]

import sklearn.metrics as sm # 模型性能评价模块
print('R2_score:', sm.r2_score(y_test,line_model(x_test,W.numpy(),B.numpy())))

R2_score: -2.399917872887695

预测效果不是很好…略微无奈，模型本身问题，解决不了房价预测问题。由此可见，选择模型非常非常重要，如果你模型选的不对，你调多好的参数都没用。