神经网络分类问题应用：中风预测&天气预测

ghostInSh3ll

已于 2023-12-19 21:16:13 修改

阅读量331

点赞数

文章标签：分类人工智能数据挖掘

于 2023-06-26 09:08:06 首次发布

本文链接：https://blog.csdn.net/AAAAshin/article/details/131389767

版权

任务1：中风预测

数据如下所示，标签stroke的取值为[0,1]，故为二分类问题。

id,gender,age,hypertension,heart_disease,ever_married,work_type,Residence_type,avg_glucose_level,bmi,smoking_status,stroke

1,Male,32,0,0,Yes,Private,Rural,82.68,29.2,never smoked,0

2,Female,50,0,0,Yes,Govt_job,Urban,104.24,32.8,Unknown,0

3,Male,15,0,0,No,Private,Rural,78.59,25.1,Unknown,0

4,Female,72,1,1,Yes,Private,Urban,198.32,31.3,formerly smoked,0

6,Female,29,0,0,No,Private,Urban,61.13,26,never smoked,0

7,Male,6,0,0,No,children,Rural,64.55,17.4,Unknown,0

9,Female,35,0,0,Yes,Private,Rural,74.55,22.4,never smoked,0

10,Female,77,1,0,Yes,Self-employed,Rural,238.53,30.6,never smoked,0

11,Female,16,0,0,No,children,Rural,109.02,19.8,Unknown,0

... ...

数据处理部分代码如下所示，

# 读取训练数据
train_data_path = r'./train.csv'
train_data = pd.read_csv(train_data_path)

# 打乱数据顺序
train_data.sample(frac=1)

# 提取真值标签
target = train_data.stroke

# 获得数据特征
train_data.drop(['stroke'], axis=1, inplace=True)

# 划分训练集和验证集
train_data, val_data, train_target, val_target = train_test_split(
    train_data, target, test_size=0.2)

# 记录训练集和验证集的数据量
num_of_train_data = train_data.shape[0]
num_of_val_data = val_data.shape[0]

# 读取测试数据
test_data_path = r'./test.csv'
test_data = pd.read_csv(test_data_path)

# 将训练集、验证集、测试集数据合并，方便之后统一进行数据处理
combined_data = train_data.append(val_data).append(test_data)
combined_data.reset_index(inplace=True)
combined_data.drop(['index', 'id'], inplace=True, axis=1)


# 缺失值处理：将bmi列缺失处替换为bmi列的均值
combined_data.fillna(value = {'bmi':combined_data['bmi'].mean()}, inplace = True)

# 处理非数值型数据，将其One Hot编码

# 获得数值型和非数值型(categorical)数据列名

def get_cols_with_no_nans(df,col_type):
    '''
    Arguments :
    df : The dataframe to process
    col_type :