【Paddle 入门打卡】用Paddle做Kaggle的波士顿房价预测

---

本文基于百度飞浆Paddle平台

项目地址：
用PaddlePaddle做房价预测

---

波士顿房价预测

---

• 框架: 百度飞浆Paddle
• 日期: 2021-11-14
• 类型: 结构化数据

---

1. 数据描述

---

波士顿数据框有 506 行和 14 列

对应特征:

• crim：犯罪率

• zn:划分为超过25,000平方英尺地段的住宅用地所占比例

• indus:每镇非零售商铺面积比例

• chas:是否临河

• nox:氮氧化物浓度(千万分之一)

• rm: 每个住宅的平均房间数

• age:一九四年以前业主自住单位比例

• dis:波士顿五个商业中心的加权平均距离

• rad:放射状公路的可达性指数

• tax：每$10,000的全价值物业税税率

• ptratio: 学生-教师比例按城镇划分

• black:1000(Bk - 0.63)^2其中Bk是按城镇划分的黑人比例

• lstat:低收入阶层人口占比

• medv：自住房屋价值中位数，以千元计

---

2.导入数据以及数据预处理

---

# 导入波士顿房价数据
import os
import paddle
import numpy as np 

# 设置训练Batch大小
BATCH_SIZE = 20

# 训练集
train_datasets = paddle.text.datasets.UCIHousing(mode= 'train')
# 验证集
valid_datasets = paddle.text.datasets.UCIHousing(mode= 'test')

# 用于训练的额数据集加载器，每次随机读取batch大小的数据，剩余不足的批次大小的数据将被丢弃
train_loader = paddle.io.DataLoader(train_datasets, batch_size=BATCH_SIZE, shuffle=True, drop_last=True)

# 测试集加载器，每次读取随机批次大小的数据
valid_loader = paddle.io.DataLoader(valid_datasets, batch_size=BATCH_SIZE, shuffle=True, drop_last=True)

# 打印数据类型
print(type(train_datasets))

<class 'paddle.text.datasets.uci_housing.UCIHousing'>

# 打印查看uci_housing数据
print(train_datasets[0])
# 每一行是一个样本，每个样本有14个特征
# print(train_datasets)

(array([-0.0405441 ,  0.06636363, -0.32356226, -0.06916996, -0.03435197,
        0.05563625, -0.03475696,  0.02682186, -0.37171334, -0.21419305,
       -0.33569506,  0.10143217, -0.21172912], dtype=float32), array([24.], dtype=float32))

---

3. 构建学习网络

---

# 定义网络结构
net = paddle.nn.Linear(13, 1)

# 定义优化函数
optimizer = paddle.optimizer.SGD(learning_rate = 0.001, parameters = net.parameters())

---

4. 模型的训练与评估

---

4.1 定义画图方法

# 定义绘图函数
import matplotlib.pyplot as plt 

iter = 0
iters = []
train_costs = []

def draw_train_process(iters, train_costs):
    title = 'training costs'
    plt.title(title, fontsize = 24)
    plt.xlabel('iter', fontsize = 14)
    plt.ylabel('cost', fontsize = 14)
    plt.plot(iters, train_costs, color = 'red', label = 'training cost')
    plt.grid()
    plt.show()
    

4.2 训练并保存模型

# 定义训练轮次
EPOCH_NUM = 50

# 训练EPOCH_NUM轮
for pass_id in range(EPOCH_NUM):
    # 开始训练并输出最后一个batch的损失值
    train_cost = 0

    # 遍历train_loader迭代器
    for batch_id, data in enumerate(train_loader()):
        # 分别提取训练集和标签
        inputs = paddle.to_tensor(data[0])
        labels = paddle.to_tensor(data[1])

        # 计算输出
        out = net(inputs)
        # 计算损失函数（均方差）
        train_loss = paddle.mean(paddle.nn.functional.square_error_cost(out, labels))

        #  反向迭代
        train_loss.backward()

        # 优化并清空dw
        optimizer.step()
        optimizer.clear_grad()

        # 每40步输出信息，
        # 从0batch开始：0, 40, 80
        if batch_id % 40 == 0:
            print("Pass id: %d, cost: %0.5f" % (pass_id, train_loss))
        
        iter = iter + BATCH_SIZE
        iters.append(iter)
        train_costs.append(train_loss.numpy()[0])

    # 开始测试并输出最后一个batch的缺失值
    test_loss = 0

    # 遍历test_reader迭代器
    for batch_id, data in enumerate(valid_loader()):
        # 分别提取训练集和标签
        inputs = paddle.to_tensor(data[0])
        labels = paddle.to_tensor(data[1])
        # 计算输出
        out = net(inputs)
        # 计算损失函数（均方差）
        train_loss = paddle.mean(paddle.nn.functional.square_error_cost(out, labels))
    
    # 打印最后一个batch的损失值 
    print("Pass id: %d, cost: %0.5f" % (pass_id, train_loss))

# 保存模型
paddle.save(net.state_dict(), 'fit_a_line.pdparams')

draw_train_process(iters, train_costs)

---

Pass id: 0, cost: 697.82361
Pass id: 0, cost: 244.72691
Pass id: 1, cost: 552.90552
Pass id: 1, cost: 221.64818
Pass id: 2, cost: 429.48627
Pass id: 2, cost: 192.41925
Pass id: 3, cost: 484.38715
Pass id: 3, cost: 173.44687
Pass id: 4, cost: 387.63263
Pass id: 4, cost: 221.82217
Pass id: 5, cost: 402.83636
Pass id: 5, cost: 128.57277
Pass id: 6, cost: 547.67670
Pass id: 6, cost: 173.92017
Pass id: 7, cost: 308.60843
Pass id: 7, cost: 147.69878
Pass id: 8, cost: 219.78215
Pass id: 8, cost: 107.03541
Pass id: 9, cost: 208.08089
Pass id: 9, cost: 112.56085
Pass id: 10, cost: 311.31699
Pass id: 10, cost: 123.64721
Pass id: 11, cost: 363.99536
Pass id: 11, cost: 105.76099
Pass id: 12, cost: 420.03595
Pass id: 12, cost: 72.11946
Pass id: 13, cost: 327.58844
Pass id: 13, cost: 88.04731
Pass id: 14, cost: 390.53604
Pass id: 14, cost: 85.30426
Pass id: 15, cost: 230.41187
Pass id: 15, cost: 63.41622
Pass id: 16, cost: 166.12924
Pass id: 16, cost: 63.62786
Pass id: 17, cost: 236.01692
Pass id: 17, cost: 77.35274
Pass id: 18, cost: 290.79498
Pass id: 18, cost: 33.24523
Pass id: 19, cost: 97.46697
Pass id: 19, cost: 51.36712
Pass id: 20, cost: 340.30243
Pass id: 20, cost: 32.06834
Pass id: 21, cost: 168.25182
Pass id: 21, cost: 38.49080
Pass id: 22, cost: 175.69730
Pass id: 22, cost: 35.95297
Pass id: 23, cost: 205.52931
Pass id: 23, cost: 71.39227
Pass id: 24, cost: 207.87970
Pass id: 24, cost: 28.49346
Pass id: 25, cost: 116.07450
Pass id: 25, cost: 35.36184
Pass id: 26, cost: 182.95099
Pass id: 26, cost: 33.28909
Pass id: 27, cost: 181.40346
Pass id: 27, cost: 40.69643
Pass id: 28, cost: 53.31166
Pass id: 28, cost: 23.08777
Pass id: 29, cost: 159.42133
Pass id: 29, cost: 49.56973
Pass id: 30, cost: 148.92783
Pass id: 30, cost: 34.58585
Pass id: 31, cost: 98.45925
Pass id: 31, cost: 22.08068
Pass id: 32, cost: 64.72221
Pass id: 32, cost: 18.82707
Pass id: 33, cost: 39.10216
Pass id: 33, cost: 41.70573
Pass id: 34, cost: 86.37087
Pass id: 34, cost: 27.22211
Pass id: 35, cost: 30.80962
Pass id: 35, cost: 32.63113
Pass id: 36, cost: 112.65536
Pass id: 36, cost: 16.17883
Pass id: 37, cost: 69.60898
Pass id: 37, cost: 19.16402
Pass id: 38, cost: 39.25970
Pass id: 38, cost: 22.66752
Pass id: 39, cost: 19.43631
Pass id: 39, cost: 24.29722
Pass id: 40, cost: 118.90907
Pass id: 40, cost: 17.93452
Pass id: 41, cost: 63.60981
Pass id: 41, cost: 21.98477
Pass id: 42, cost: 48.37024
Pass id: 42, cost: 42.80917
Pass id: 43, cost: 120.80444
Pass id: 43, cost: 16.80975
Pass id: 44, cost: 67.58469
Pass id: 44, cost: 19.20858
Pass id: 45, cost: 19.36415
Pass id: 45, cost: 31.67330
Pass id: 46, cost: 67.76525
Pass id: 46, cost: 19.26246
Pass id: 47, cost: 91.39543
Pass id: 47, cost: 30.07190
Pass id: 48, cost: 120.14481
Pass id: 48, cost: 17.26798
Pass id: 49, cost: 89.36503
Pass id: 49, cost: 21.93978


/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2349: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working
  if isinstance(obj, collections.Iterator):
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2366: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working
  return list(data) if isinstance(data, collections.MappingView) else data

---

5. 模型预测

---

5.1 模型的可视化真实值与预测值方法定义

infer_results = []
groud_truths = []

# 绘制真实值和预测值对比图
def draw_infer_result(groud_truths, infer_results):
    title = 'Boston'
    plt.title(title, fontsize = 24)
    x = np.arange(1, 20)
    y = x
    plt.plot(x, y)
    plt.xlabel('ground truth', fontsize = 14)
    plt.ylabel('infer result', fontsize = 14)
    plt.scatter(groud_truths, infer_results, color = 'green', label = 'training costs')
    plt.legend()
    plt.grid()
    plt.show()

5.2 预测模型

import paddle
import numpy as np 
import matplotlib.pyplot as plt

valid_datasets = paddle.text.UCIHousing(mode='test')
infer_loader = paddle.io.DataLoader(valid_datasets, batch_size=200)

# 先创建一个架构
infer_net = paddle.nn.Linear(13, 1)
# 加载模型
param = paddle.load('fit_a_line.pdparams')

infer_net.set_dict(param)

data = next(infer_loader())
inputs  = paddle.to_tensor(data[0])
results = infer_net(inputs)

for idx, item in enumerate(zip(results, data[1])):
    print("Index:%d, Infer Result: %.2f, Ground Truth: %.2f" % (idx, item[0], item[1]))
    infer_results.append(item[0].numpy()[0])
    groud_truths.append(item[1].numpy()[0])

draw_infer_result(groud_truths, infer_results)

Index:0, Infer Result: 12.86, Ground Truth: 8.50
Index:1, Infer Result: 12.75, Ground Truth: 5.00
Index:2, Infer Result: 12.67, Ground Truth: 11.90
Index:3, Infer Result: 14.19, Ground Truth: 27.90
Index:4, Infer Result: 13.26, Ground Truth: 17.20
Index:5, Infer Result: 13.88, Ground Truth: 27.50
Index:6, Infer Result: 13.41, Ground Truth: 15.00
Index:7, Infer Result: 13.40, Ground Truth: 17.20
Index:8, Infer Result: 11.47, Ground Truth: 17.90
Index:9, Infer Result: 13.08, Ground Truth: 16.30
Index:10, Infer Result: 10.83, Ground Truth: 7.00
Index:11, Infer Result: 12.48, Ground Truth: 7.20
Index:12, Infer Result: 13.14, Ground Truth: 7.50
Index:13, Infer Result: 12.54, Ground Truth: 10.40
Index:14, Infer Result: 12.28, Ground Truth: 8.80
Index:15, Infer Result: 13.67, Ground Truth: 8.40
Index:16, Infer Result: 14.19, Ground Truth: 16.70
Index:17, Infer Result: 14.12, Ground Truth: 14.20
Index:18, Infer Result: 14.37, Ground Truth: 20.80
Index:19, Infer Result: 13.34, Ground Truth: 13.40
Index:20, Infer Result: 13.96, Ground Truth: 11.70
Index:21, Infer Result: 12.72, Ground Truth: 8.30
Index:22, Infer Result: 14.42, Ground Truth: 10.20
Index:23, Infer Result: 13.74, Ground Truth: 10.90
Index:24, Infer Result: 13.72, Ground Truth: 11.00
Index:25, Infer Result: 13.13, Ground Truth: 9.50
Index:26, Infer Result: 14.08, Ground Truth: 14.50
Index:27, Infer Result: 13.92, Ground Truth: 14.10
Index:28, Infer Result: 14.86, Ground Truth: 16.10
Index:29, Infer Result: 14.02, Ground Truth: 14.30
Index:30, Infer Result: 13.76, Ground Truth: 11.70
Index:31, Infer Result: 13.28, Ground Truth: 13.40
Index:32, Infer Result: 13.44, Ground Truth: 9.60
Index:33, Infer Result: 12.45, Ground Truth: 8.70
Index:34, Infer Result: 12.15, Ground Truth: 8.40
Index:35, Infer Result: 13.51, Ground Truth: 12.80
Index:36, Infer Result: 13.52, Ground Truth: 10.50
Index:37, Infer Result: 14.01, Ground Truth: 17.10
Index:38, Infer Result: 14.17, Ground Truth: 18.40
Index:39, Infer Result: 14.03, Ground Truth: 15.40
Index:40, Infer Result: 13.11, Ground Truth: 10.80
Index:41, Infer Result: 13.01, Ground Truth: 11.80
Index:42, Infer Result: 14.04, Ground Truth: 14.90
Index:43, Infer Result: 14.22, Ground Truth: 12.60
Index:44, Infer Result: 14.10, Ground Truth: 14.10
Index:45, Infer Result: 13.93, Ground Truth: 13.00
Index:46, Infer Result: 13.73, Ground Truth: 13.40
Index:47, Infer Result: 14.30, Ground Truth: 15.20
Index:48, Infer Result: 14.39, Ground Truth: 16.10
Index:49, Infer Result: 14.67, Ground Truth: 17.80
Index:50, Infer Result: 13.58, Ground Truth: 14.90
Index:51, Infer Result: 13.85, Ground Truth: 14.10
Index:52, Infer Result: 13.45, Ground Truth: 12.70
Index:53, Infer Result: 13.72, Ground Truth: 13.50
Index:54, Infer Result: 14.41, Ground Truth: 14.90
Index:55, Infer Result: 14.69, Ground Truth: 20.00
Index:56, Infer Result: 14.41, Ground Truth: 16.40
Index:57, Infer Result: 14.75, Ground Truth: 17.70
Index:58, Infer Result: 14.88, Ground Truth: 19.50
Index:59, Infer Result: 15.11, Ground Truth: 20.20
Index:60, Infer Result: 15.39, Ground Truth: 21.40
Index:61, Infer Result: 15.43, Ground Truth: 19.90
Index:62, Infer Result: 13.83, Ground Truth: 19.00
Index:63, Infer Result: 14.07, Ground Truth: 19.10
Index:64, Infer Result: 14.76, Ground Truth: 19.10
Index:65, Infer Result: 15.33, Ground Truth: 20.10
Index:66, Infer Result: 14.93, Ground Truth: 19.90
Index:67, Infer Result: 15.19, Ground Truth: 19.60
Index:68, Infer Result: 15.38, Ground Truth: 23.20
Index:69, Infer Result: 15.82, Ground Truth: 29.80
Index:70, Infer Result: 14.06, Ground Truth: 13.80
Index:71, Infer Result: 13.74, Ground Truth: 13.30
Index:72, Infer Result: 14.54, Ground Truth: 16.70
Index:73, Infer Result: 13.26, Ground Truth: 12.00
Index:74, Infer Result: 14.30, Ground Truth: 14.60
Index:75, Infer Result: 14.83, Ground Truth: 21.40
Index:76, Infer Result: 15.92, Ground Truth: 23.00
Index:77, Infer Result: 16.14, Ground Truth: 23.70
Index:78, Infer Result: 16.29, Ground Truth: 25.00
Index:79, Infer Result: 16.34, Ground Truth: 21.80
Index:80, Infer Result: 15.95, Ground Truth: 20.60
Index:81, Infer Result: 16.18, Ground Truth: 21.20
Index:82, Infer Result: 15.11, Ground Truth: 19.10
Index:83, Infer Result: 15.84, Ground Truth: 20.60
Index:84, Infer Result: 15.65, Ground Truth: 15.20
Index:85, Infer Result: 14.94, Ground Truth: 7.00
Index:86, Infer Result: 14.31, Ground Truth: 8.10
Index:87, Infer Result: 15.73, Ground Truth: 13.60
Index:88, Infer Result: 16.46, Ground Truth: 20.10
Index:89, Infer Result: 20.31, Ground Truth: 21.80
Index:90, Infer Result: 20.51, Ground Truth: 24.50
Index:91, Infer Result: 20.40, Ground Truth: 23.10
Index:92, Infer Result: 19.09, Ground Truth: 19.70
Index:93, Infer Result: 19.87, Ground Truth: 18.30
Index:94, Infer Result: 20.13, Ground Truth: 21.20
Index:95, Infer Result: 19.60, Ground Truth: 17.50
Index:96, Infer Result: 19.72, Ground Truth: 16.80
Index:97, Infer Result: 21.10, Ground Truth: 22.40
Index:98, Infer Result: 20.79, Ground Truth: 20.60
Index:99, Infer Result: 21.10, Ground Truth: 23.90
Index:100, Infer Result: 21.01, Ground Truth: 22.00
Index:101, Infer Result: 20.78, Ground Truth: 11.90

---

写在最后

各位看官，都看到这里了，麻烦动动手指头给博主来个点赞8，您的支持作者最大的创作动力哟！
＜（＾－＾）＞
才疏学浅，若有纰漏，恳请斧正
本文章仅用于各位同志作为学习交流之用，不作任何商业用途，若涉及版权问题请速与作者联系，望悉知