import torch
import torch.nn as nn
import torch.optim as optim
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler
import time
import matplotlib.pyplot as plt
from tqdm import tqdm  # 导入tqdm库用于进度条显示
import warnings
warnings.filterwarnings("ignore")  # 忽略警告信息
 
# 设置GPU设备
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print(f"使用设备: {device}")
 
import pandas as pd
# 加载信贷数据集
data = pd.read_csv(r"E:\PythonStudy\python60-days-challenge-master\data.csv")
 
import numpy as np     #用于数值计算，提供了高效的数组操作。
# 先筛选字符串变量 
discrete_features = data.select_dtypes(include=['object']).columns.tolist()
# Home Ownership 标签编码
home_ownership_mapping = {
    'Own Home': 1,
    'Rent': 2,
    'Have Mortgage': 3,
    'Home Mortgage': 4
}
data['Home Ownership'] = data['Home Ownership'].map(home_ownership_mapping)
 
# Years in current job 标签编码
years_in_job_mapping = {
    '< 1 year': 1,
    '1 year': 2,
    '2 years': 3,
    '3 years': 4,
    '4 years': 5,
    '5 years': 6,
    '6 years': 7,
    '7 years': 8,
    '8 years': 9,
    '9 years': 10,
    '10+ years': 11
}
data['Years in current job'] = data['Years in current job'].map(years_in_job_mapping)
 
# Purpose 独热编码，记得需要将bool类型转换为数值
data = pd.get_dummies(data, columns=['Purpose'])
data2 = pd.read_csv(r"E:\PythonStudy\python60-days-challenge-master\data.csv") # 重新读取数据，用来做列名对比
list_final = [] # 新建一个空列表，用于存放独热编码后新增的特征名
for i in data.columns:
    if i not in data2.columns:
       list_final.append(i) # 这里打印出来的就是独热编码后的特征名
for i in list_final:
    data[i] = data[i].astype(int) # 这里的i就是独热编码后的特征名
 
 
 
# Term 0 - 1 映射
term_mapping = {
    'Short Term': 0,
    'Long Term': 1
}
data['Term'] = data['Term'].map(term_mapping)
data.rename(columns={'Term': 'Long Term'}, inplace=True) # 重命名列
continuous_features = data.select_dtypes(include=['int64', 'float64']).columns.tolist()  #把筛选出来的列名转换成列表
 
 # 连续特征用中位数补全
for feature in continuous_features:     
    mode_value = data[feature].mode()[0]            #获取该列的众数。
    data[feature].fillna(mode_value, inplace=True)          #用众数填充该列的缺失值，inplace=True表示直接在原数据上修改。
 
# 最开始也说了 很多调参函数自带交叉验证，甚至是必选的参数，你如果想要不交叉反而实现起来会麻烦很多
# 所以这里我们还是只划分一次数据集
from sklearn.model_selection import train_test_split
X = data.drop(['Credit Default'], axis=1)  # 特征，axis=1表示按列删除
y = data['Credit Default'] # 标签
# 按照7:3划分训练集和测试集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)  
#将测试集1：1划分验证集和测试集
X_val, X_test, y_val, y_test = train_test_split(X_test, y_test, test_size=0.5, random_state=42)  # 50%验证集，50%测试集
#归一化处理
scaler = MinMaxScaler()
X_train = scaler.fit_transform(X_train) 
X_val = scaler.transform(X_val)
X_test = scaler.transform(X_test)
X_train = torch.tensor(X_train, dtype=torch.float32).to(device)
X_val = torch.tensor(X_val, dtype=torch.float32).to(device)
X_test = torch.tensor(X_test, dtype=torch.float32).to(device)
y_train = torch.tensor(y_train.values, dtype=torch.float32).view(-1, 1).to(device)
y_val = torch.tensor(y_val.values, dtype=torch.float32).view(-1, 1).to(device)
y_test = torch.tensor(y_test.values, dtype=torch.float32).view(-1, 1).to(device)
 
#定义神经网络模型
# 定义模型（简单的3层全连接网络，相当于“小炒锅”，足够炒信贷数据这个菜）
import torch.nn as nn
 
class CreditModel(nn.Module):
    def __init__(self, input_dim):
        super(CreditModel, self).__init__()
        # 网络层：输入层→隐藏层1→隐藏层2→输出层（二分类只有1个输出）
        self.fc1 = nn.Linear(input_dim, 128)  # 第一层：输入维度→128个神经元
        self.fc2 = nn.Linear(128, 64)         # 第二层：128→64
        self.fc3 = nn.Linear(64, 1)           # 输出层：64→1（预测是否违约）
        self.relu = nn.ReLU()                 # 激活函数（让模型学复杂规律）
        self.sigmoid = nn.Sigmoid()           # 输出0~1之间的概率
 
    def forward(self, x):
        # 前向传播（数据在网络里的流动）
        x = self.relu(self.fc1(x))
        x = self.relu(self.fc2(x))
        x = self.sigmoid(self.fc3(x))
        return x
 
# 实例化模型（把锅架起来）
model = CreditModel(X_train.shape[1]).to(device)
 
# 定义损失函数和优化器
criterion = nn.BCELoss()  # 二分类任务的损失函数（适合判断违约）
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)  # 优化器（让模型学的更快）
 
# 构建数据加载器
from torch.utils.data import TensorDataset, DataLoader
 
batch_size = 32
train_dataset = TensorDataset(X_train, y_train)
val_dataset = TensorDataset(X_val, y_val)
test_dataset = TensorDataset(X_test, y_test)
 
train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False)
test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False)
 
# 查看处理后的数据维度（确认没问题）
input_dim = X_train.shape[1]
print(f"模型输入维度：{input_dim}")
 
 
@浙大疏锦行