错误修改系列---基于RNN模型的心脏病预测(pytorch实现)

前言

• 前几天发布了pytorch实现，TensorFlow实现为：基于RNN模型的心脏病预测(tensorflow实现)，但是一处繁琐地方 + 一处错误，这篇文章进行修改，修改效果还是好了不少；
• 源文章为：基于RNN模型的心脏病预测，提供tensorflow和pytorch实现

---

错误一
这个也不算是错误，就是之前数据标准化、划分数据集的时候，我用的很麻烦，如下图(之前):

这样无疑是很麻烦的，修改后，我们先对数据进行标准化，后再进行划分就会简单很多(详细请看下面代码)

---

错误二
模型参数输入，这里应该是13个特征维度，而且这里用nn.BCELoss后面处理也不好，因为最后应该还加一层激活函数sigmoid的，所以这次修改采用多分类处理方法，激活函数采用CrossEntropyLoss，具体如图：

BCELoss、CrossEntropyLoss参考资料：

https://blog.csdn.net/qq_36803941/article/details/138673111
https://zhuanlan.zhihu.com/p/98785902
https://www.cnblogs.com/zhangxianrong/p/14773075.html
https://zhuanlan.zhihu.com/p/59800597

修改版本代码

1、数据处理

1、导入库

import pandas as pd 
import numpy as np 
import matplotlib.pyplot as plt 
from torch.utils.data import DataLoader, TensorDataset
import torch device = 'cuda' if torch.cuda.is_available() else 'cpu'
device

'cuda'

2、导入数据

data = pd.read_csv('./heart.csv')data.head()

	age	sex	cp	trestbps	chol	fbs	restecg	thalach	exang	oldpeak	slope	ca	thal	target
0	63	1	3	145	233	1	0	150	0	2.3	0	0	1	1
1	37	1	2	130	250	0	1	187	0	3.5	0	0	2	1
2	41	0	1	130	204	0	0	172	0	1.4	2	0	2	1
3	56	1	1	120	236	0	1	178	0	0.8	2	0	2	1
4	57	0	0	120	354	0	1	163	1	0.6	2	0	2	1

• age - 年龄
• sex - (1 = male(男性); 0 = (女性))
• cp - chest pain type(胸部疼痛类型)（1：典型的心绞痛-typical，2：非典型心绞痛-atypical，3：没有心绞痛-non-anginal，4：无症状-asymptomatic）
• trestbps - 静息血压 (in mm Hg on admission to the hospital)
• chol - 胆固醇 in mg/dl
• fbs - (空腹血糖 > 120 mg/dl) (1 = true; 0 = false)
• restecg - 静息心电图测量（0：普通，1：ST-T波异常，2：可能左心室肥大）
• thalach - 最高心跳率
• exang - 运动诱发心绞痛 (1 = yes; 0 = no)
• oldpeak - 运动相对于休息引起的ST抑制
• slope - 运动ST段的峰值斜率（1：上坡-upsloping，2：平的-flat，3：下坡-downsloping）
• ca - 主要血管数目(0-4)
• thal - 一种叫做地中海贫血的血液疾病（3 = normal; 6 = 固定的缺陷-fixed defect; 7 = 可逆的缺陷-reversable defect）
• target - 是否患病 (1=yes, 0=no)

3、数据分析

数据初步分析

data.info()   # 数据类型分析

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 303 entries, 0 to 302
Data columns (total 14 columns):#   Column    Non-Null Count  Dtype  
---  ------    --------------  -----  0   age       303 non-null    int64  1   sex       303 non-null    int64  2   cp        303 non-null    int64  3   trestbps  303 non-null    int64  4   chol      303 non-null    int64  5   fbs       303 non-null    int64  6   restecg   303 non-null    int64  7   thalach   303 non-null    int64  8   exang     303 non-null    int64  9   oldpeak   303 non-null    float6410  slope     303 non-null    int64  11  ca        303 non-null    int64  12  thal      303 non-null    int64  13  target    303 non-null    int64  
dtypes: float64(1), int64(13)
memory usage: 33.3 KB

其中分类变量为：sex、cp、fbs、restecg、exang、slope、ca、thal、target

数值型变量：age、trestbps、chol、thalach、oldpeak

data.describe()  # 描述性

	age	sex	cp	trestbps	chol	fbs	restecg	thalach	exang	oldpeak	slope	ca	thal	target
count	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000	303.000000
mean	54.366337	0.683168	0.966997	131.623762	246.264026	0.148515	0.528053	149.646865	0.326733	1.039604	1.399340	0.729373	2.313531	0.544554
std	9.082101	0.466011	1.032052	17.538143	51.830751	0.356198	0.525860	22.905161	0.469794	1.161075	0.616226	1.022606	0.612277	0.498835
min	29.000000	0.000000	0.000000	94.000000	126.000000	0.000000	0.000000	71.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	47.500000	0.000000	0.000000	120.000000	211.000000	0.000000	0.000000	133.500000	0.000000	0.000000	1.000000	0.000000	2.000000	0.000000
50%	55.000000	1.000000	1.000000	130.000000	240.000000	0.000000	1.000000	153.000000	0.000000	0.800000	1.000000	0.000000	2.000000	1.000000
75%	61.000000	1.000000	2.000000	140.000000	274.500000	0.000000	1.000000	166.000000	1.000000	1.600000	2.000000	1.000000	3.000000	1.000000
max	77.000000	1.000000	3.000000	200.000000	564.000000	1.000000	2.000000	202.000000	1.000000	6.200000	2.000000	4.000000	3.000000	1.000000

• 年纪：均值54，中位数55，标准差9，说明主要是老年人，偏大
• 静息血压：均值131.62， 成年人一般：正常血压：收缩压 < 120 mmHg，偏大
• 胆固醇：均值246.26，理想水平：小于 200 mg/dL，偏大
• 最高心率：均值149.64，一般静息状态下通常是 60 到 100 次每分钟，偏大

最大值和最小值都可能发生，无异常值

缺失值

data.isnull().sum()

age         0
sex         0
cp          0
trestbps    0
chol        0
fbs         0
restecg     0
thalach     0
exang       0
oldpeak     0
slope       0
ca          0
thal        0
target      0
dtype: int64

相关性分析

import seaborn as snsplt.figure(figsize=(20, 15))sns.heatmap(data.corr(), annot=True, cmap='Greens')plt.show()

​

​

相关系数的等级划分

• 非常弱的相关性：
  • 0.00 至 0.19 或 -0.00 至 -0.19
  • 解释：几乎不存在线性关系。
• 弱相关性：
  • 0.20 至 0.39 或 -0.20 至 -0.39
  • 解释：存在一定的线性关系，但较弱。
• 中等相关性：
  • 0.40 至 0.59 或 -0.40 至 -0.59
  • 解释：有明显的线性关系，但不是特别强。
• 强相关性：
  • 0.60 至 0.79 或 -0.60 至 -0.79
  • 解释：两个变量之间有较强的线性关系。
• 非常强的相关性：
  • 0.80 至 1.00 或 -0.80 至 -1.00
  • 解释：几乎完全线性相关，表明两个变量的变化高度一致。

---

target与chol、没有什么相关性，fbs是分类变量，chol胆固醇是数值型变量，但是从实际角度，这些都有影响，故不剔除特征

4、数据标准化

from sklearn.preprocessing import StandardScalerscaler = StandardScaler()X = data.iloc[:, :-1]
y = data.iloc[:, -1]# 这里只需要对X标准化即可
X = scaler.fit_transform(X)

5、数据划分

这里先划分为：训练集：测试集 = 9：1

from sklearn.model_selection import train_test_split# 由于要使用pytorch，先将数据转化为torch
X = torch.tensor(np.array(X), dtype=torch.float32)
y = torch.tensor(np.array(y), dtype=torch.int64)X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=42)# 输出维度
X_train.shape, y_train.shape

(torch.Size([272, 13]), torch.Size([272]))

6、动态加载数据

from torch.utils.data import TensorDataset, DataLoader 
train_dl = DataLoader(TensorDataset(X_train, y_train),                       batch_size=64,                        shuffle=True) 
test_dl  = DataLoader(TensorDataset(X_test, y_test),                       batch_size=64,                        shuffle=False)

2、创建模型

• 定义一个RNN层
  rnn = nn.RNN(input_size=10, hidden_size=20, num_layers=2, nonlinearity=‘tanh’,
  bias=True, batch_first=False, dropout=0, bidirectional=False)
• input_size: 输入的特征维度
• hidden_size: 隐藏层的特征维度
• num_layers: RNN 层的数量
• nonlinearity: 非线性激活函数 (‘tanh’ 或 ‘relu’)
• bias: 如果为 False，则内部不含偏置项，默认为 True
• batch_first: 如果为 True，则输入和输出张量提供为 (batch, seq, feature)，默认为 False (seq, batch, feature)
• dropout: 如果非零，则除了最后一层，在每层的输出中引入一个 Dropout 层，默认为 0
• bidirectional: 如果为 True，则将成为双向 RNN，默认为 False

import torch  
import torch.nn as nn # 创建模型
'''
该问题本质是二分类问题，故最后一层全连接层用激活函数为：sigmoid
模型结构：RNN：隐藏层200，激活函数：reluLinear：--> 100(relu) -> 1(sigmoid)
'''
# 创建模型
class Model(nn.Module):def __init__(self):super().__init__()self.rnn = nn.RNN(input_size=13, hidden_size=200, num_layers=1, batch_first=True)self.fc1 = nn.Linear(200, 50)#self.fc2 = nn.Linear(100, 50)self.fc3 = nn.Linear(50, 2)def forward(self, x):x, hidden1 = self.rnn(x)x = self.fc1(x)#x = self.fc2(x)x = self.fc3(x)return xmodel = Model().to(device)
model

Model((rnn): RNN(13, 200, batch_first=True)(fc1): Linear(in_features=200, out_features=50, bias=True)(fc3): Linear(in_features=50, out_features=2, bias=True)
)

# 查看模型输出的维度
model(torch.rand(30,13).to(device)).shape

torch.Size([30, 2])

3、模型训练

1、设置超参数

loss_fn = nn.CrossEntropyLoss()
lr = 1e-4
optimizer = torch.optim.Adam(model.parameters(), lr=lr)

2、设置训练函数

def train(dataloader, model, loss_fn, optimizer):# 总大小size = len(dataloader.dataset)# 批次大小batch_size = len(dataloader)# 准确率和损失trian_acc, train_loss = 0, 0# 训练for X, y in dataloader:X, y = X.to(device), y.to(device)# 模型训练与误差评分pred = model(X)loss = loss_fn(pred, y)# 梯度清零optimizer.zero_grad()  # 梯度上更新# 方向传播loss.backward()# 梯度更新optimizer.step()# 记录损失和准确率train_loss += loss.item()trian_acc += (pred.argmax(1) == y).type(torch.float64).sum().item()# 计算损失和准确率trian_acc /= sizetrain_loss /= batch_sizereturn trian_acc, train_loss

3、设置测试函数

def test(dataloader, model, loss_fn):size = len(dataloader.dataset)batch_size = len(dataloader)test_acc, test_loss = 0, 0with torch.no_grad():for X, y in dataloader:X, y = X.to(device), y.to(device)pred = model(X)loss = loss_fn(pred, y)test_loss += loss.item()test_acc += (pred.argmax(1) == y).type(torch.float64).sum().item()test_acc /= size test_loss /= batch_sizereturn test_acc, test_loss

4、模型训练

train_acc = []
train_loss = []
test_acc = []
test_loss = []# 定义训练次数
epoches = 50for epoch in range(epoches):# 训练model.train()epoch_trian_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)# 测试model.eval()epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)# 记录train_acc.append(epoch_trian_acc)train_loss.append(epoch_train_loss)test_acc.append(epoch_test_acc)test_loss.append(epoch_test_loss)template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%, Test_loss:{:.3f}')print(template.format(epoch+1, epoch_trian_acc*100, epoch_train_loss, epoch_test_acc*100, epoch_test_loss))

Epoch: 1, Train_acc:49.6%, Train_loss:0.686, Test_acc:58.1%, Test_loss:0.684
Epoch: 2, Train_acc:62.1%, Train_loss:0.682, Test_acc:64.5%, Test_loss:0.671
Epoch: 3, Train_acc:68.0%, Train_loss:0.662, Test_acc:71.0%, Test_loss:0.658
Epoch: 4, Train_acc:69.1%, Train_loss:0.655, Test_acc:77.4%, Test_loss:0.645
Epoch: 5, Train_acc:73.9%, Train_loss:0.643, Test_acc:80.6%, Test_loss:0.632
Epoch: 6, Train_acc:74.3%, Train_loss:0.637, Test_acc:80.6%, Test_loss:0.620
Epoch: 7, Train_acc:75.7%, Train_loss:0.620, Test_acc:80.6%, Test_loss:0.608
Epoch: 8, Train_acc:78.3%, Train_loss:0.612, Test_acc:80.6%, Test_loss:0.596
Epoch: 9, Train_acc:79.8%, Train_loss:0.591, Test_acc:83.9%, Test_loss:0.586
Epoch:10, Train_acc:79.0%, Train_loss:0.590, Test_acc:83.9%, Test_loss:0.575
Epoch:11, Train_acc:81.2%, Train_loss:0.584, Test_acc:83.9%, Test_loss:0.563
Epoch:12, Train_acc:79.8%, Train_loss:0.562, Test_acc:83.9%, Test_loss:0.553
Epoch:13, Train_acc:80.5%, Train_loss:0.546, Test_acc:83.9%, Test_loss:0.542
Epoch:14, Train_acc:80.1%, Train_loss:0.546, Test_acc:83.9%, Test_loss:0.531
Epoch:15, Train_acc:81.2%, Train_loss:0.517, Test_acc:83.9%, Test_loss:0.521
Epoch:16, Train_acc:81.6%, Train_loss:0.521, Test_acc:83.9%, Test_loss:0.509
Epoch:17, Train_acc:82.4%, Train_loss:0.508, Test_acc:83.9%, Test_loss:0.497
Epoch:18, Train_acc:82.7%, Train_loss:0.494, Test_acc:83.9%, Test_loss:0.487
Epoch:19, Train_acc:83.1%, Train_loss:0.496, Test_acc:83.9%, Test_loss:0.477
Epoch:20, Train_acc:82.4%, Train_loss:0.469, Test_acc:83.9%, Test_loss:0.469
Epoch:21, Train_acc:83.1%, Train_loss:0.472, Test_acc:83.9%, Test_loss:0.463
Epoch:22, Train_acc:82.4%, Train_loss:0.451, Test_acc:83.9%, Test_loss:0.458
Epoch:23, Train_acc:83.5%, Train_loss:0.456, Test_acc:83.9%, Test_loss:0.455
Epoch:24, Train_acc:83.1%, Train_loss:0.438, Test_acc:83.9%, Test_loss:0.453
Epoch:25, Train_acc:83.5%, Train_loss:0.431, Test_acc:80.6%, Test_loss:0.451
Epoch:26, Train_acc:84.2%, Train_loss:0.444, Test_acc:80.6%, Test_loss:0.449
Epoch:27, Train_acc:83.1%, Train_loss:0.427, Test_acc:80.6%, Test_loss:0.449
Epoch:28, Train_acc:84.2%, Train_loss:0.409, Test_acc:80.6%, Test_loss:0.449
Epoch:29, Train_acc:83.8%, Train_loss:0.405, Test_acc:80.6%, Test_loss:0.448
Epoch:30, Train_acc:83.8%, Train_loss:0.411, Test_acc:80.6%, Test_loss:0.448
Epoch:31, Train_acc:83.8%, Train_loss:0.378, Test_acc:80.6%, Test_loss:0.446
Epoch:32, Train_acc:84.6%, Train_loss:0.421, Test_acc:80.6%, Test_loss:0.444
Epoch:33, Train_acc:84.6%, Train_loss:0.391, Test_acc:80.6%, Test_loss:0.443
Epoch:34, Train_acc:85.7%, Train_loss:0.388, Test_acc:80.6%, Test_loss:0.446
Epoch:35, Train_acc:84.2%, Train_loss:0.396, Test_acc:80.6%, Test_loss:0.449
Epoch:36, Train_acc:84.2%, Train_loss:0.346, Test_acc:80.6%, Test_loss:0.451
Epoch:37, Train_acc:84.9%, Train_loss:0.379, Test_acc:80.6%, Test_loss:0.453
Epoch:38, Train_acc:84.9%, Train_loss:0.389, Test_acc:80.6%, Test_loss:0.453
Epoch:39, Train_acc:83.1%, Train_loss:0.386, Test_acc:80.6%, Test_loss:0.453
Epoch:40, Train_acc:84.9%, Train_loss:0.350, Test_acc:80.6%, Test_loss:0.452
Epoch:41, Train_acc:83.5%, Train_loss:0.353, Test_acc:80.6%, Test_loss:0.455
Epoch:42, Train_acc:85.7%, Train_loss:0.373, Test_acc:80.6%, Test_loss:0.458
Epoch:43, Train_acc:84.6%, Train_loss:0.345, Test_acc:80.6%, Test_loss:0.459
Epoch:44, Train_acc:85.3%, Train_loss:0.377, Test_acc:80.6%, Test_loss:0.461
Epoch:45, Train_acc:85.7%, Train_loss:0.354, Test_acc:80.6%, Test_loss:0.462
Epoch:46, Train_acc:84.9%, Train_loss:0.327, Test_acc:80.6%, Test_loss:0.467
Epoch:47, Train_acc:82.7%, Train_loss:0.347, Test_acc:80.6%, Test_loss:0.470
Epoch:48, Train_acc:84.6%, Train_loss:0.350, Test_acc:80.6%, Test_loss:0.470
Epoch:49, Train_acc:84.9%, Train_loss:0.344, Test_acc:80.6%, Test_loss:0.470
Epoch:50, Train_acc:85.3%, Train_loss:0.375, Test_acc:80.6%, Test_loss:0.472

5、结果展示

import matplotlib.pyplot as plt
#隐藏警告
import warnings
warnings.filterwarnings("ignore")               #忽略警告信息
plt.rcParams['font.sans-serif']    = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False      # 用来正常显示负号
plt.rcParams['figure.dpi']         = 100        #分辨率epoch_length = range(epoches)plt.figure(figsize=(12, 3))plt.subplot(1, 2, 1)
plt.plot(epoch_length, train_acc, label='Train Accuaray')
plt.plot(epoch_length, test_acc, label='Test Accuaray')
plt.legend(loc='lower right')
plt.title('Accurary')plt.subplot(1, 2, 2)
plt.plot(epoch_length, train_loss, label='Train Loss')
plt.plot(epoch_length, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Loss')plt.show()

​

​

趋于平稳不是没有变化，是变化很小，整体模型效果还可以

6、模型评估

# 评估：返回的是自己在model.compile中设置，这里为accuracy
test_acc, test_loss = test(test_dl, model, loss_fn)
print("socre[loss, accuracy]: ", test_acc, test_loss) # 返回为两个，一个是loss，一个是accuracy

socre[loss, accuracy]:  0.8064516129032258 0.47150832414627075