day40复习日@浙大疏锦行
day40复习日@浙大疏锦行
用现有的知识对信贷风险预测数据集建立模型进行了训练
importpandasaspdimportnumpyasnpimporttorchimporttorch.nnasnnimporttorch.optimasoptimfromsklearn.model_selectionimporttrain_test_splitfromsklearn.preprocessingimportStandardScaler,LabelEncoderfromsklearn.imputeimportSimpleImputerimportmatplotlib.pyplotaspltfromtqdmimporttqdmimporttime# 设置随机种子torch.manual_seed(42)np.random.seed(42)# 设置GPU设备device=torch.device("cuda:0"iftorch.cuda.is_available()else"cpu")print(f"使用设备:{device}")# 1. 读取数据# 注意：根据文件结构，data.csv 在上一级目录df=pd.read_csv('../data.csv')# 2. 数据预处理# 删除ID列，因为它对预测没有帮助if'Id'indf.columns:df=df.drop('Id',axis=1)# 分离特征和标签X=df.drop('Credit Default',axis=1)y=df['Credit Default']# 区分数值型和类别型特征numerical_cols=X.select_dtypes(include=['int64','float64']).columns categorical_cols=X.select_dtypes(include=['object']).columns# 缺失值处理# 数值型用均值填充imputer_num=SimpleImputer(strategy='mean')X_num=pd.DataFrame(imputer_num.fit_transform(X[numerical_cols]),columns=numerical_cols)# 类别型用众数填充imputer_cat=SimpleImputer(strategy='most_frequent')X_cat=pd.DataFrame(imputer_cat.fit_transform(X[categorical_cols]),columns=categorical_cols)# 类别特征编码 (使用One-Hot编码)X_cat=pd.get_dummies(X_cat,drop_first=True)# 合并数据X=pd.concat([X_num,X_cat],axis=1)# 数据标准化scaler=StandardScaler()X=scaler.fit_transform(X)# 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=0.2,random_state=42)# 4. 转换为Tensor并移至GPUX_train=torch.FloatTensor(X_train).to(device)y_train=torch.LongTensor(y_train.values).to(device)X_test=torch.FloatTensor(X_test).to(device)y_test=torch.LongTensor(y_test.values).to(device)print(f"训练集形状:{X_train.shape}")print(f"测试集形状:{X_test.shape}")训练集形状: torch.Size([6000, 40])
测试集形状: torch.Size([1500, 40])
测试集形状: torch.Size([1500, 40])
classCreditModel(nn.Module):def__init__(self,input_dim):super(CreditModel,self).__init__()self.fc1=nn.Linear(input_dim,128)self.relu1=nn.ReLU()self.dropout1=nn.Dropout(0.3)self.fc2=nn.Linear(128,64)self.relu2=nn.ReLU()self.dropout2=nn.Dropout(0.3)self.fc3=nn.Linear(64,2)# 二分类输出defforward(self,x):out=self.fc1(x)out=self.relu1(out)out=self.dropout1(out)out=self.fc2(out)out=self.relu2(out)out=self.dropout2(out)out=self.fc3(out)returnout# 获取输入维度input_dim=X_train.shape[1]model=CreditModel(input_dim).to(device)print(model)# 定义损失函数和优化器criterion=nn.CrossEntropyLoss()optimizer=optim.Adam(model.parameters(),lr=0.001)# 训练参数num_epochs=1000losses=[]start_time=time.time()# 使用tqdm显示进度条withtqdm(total=num_epochs,desc="训练进度",unit="epoch")aspbar:forepochinrange(num_epochs):model.train()# 确保在训练模式 (因为有Dropout)# 前向传播outputs=model(X_train)loss=criterion(outputs,y_train)# 反向传播和优化optimizer.zero_grad()loss.backward()optimizer.step()# 记录损失losses.append(loss.item())# 更新进度条if(epoch+1)%10==0:pbar.set_postfix({'Loss':f'{loss.item():.4f}'})pbar.update(10)print(f"训练耗时:{time.time()-start_time:.2f}秒")# 绘制损失曲线plt.figure(figsize=(10,5))plt.plot(losses)plt.title('Training Loss')plt.xlabel('Epoch')plt.ylabel('Loss')plt.show()# 模型评估model.eval()# 切换到评估模式withtorch.no_grad():outputs=model(X_test)_,predicted=torch.max(outputs.data,1)total=y_test.size(0)correct=(predicted==y_test).sum().item()accuracy=100*correct/totalprint(f'测试集准确率:{accuracy:.2f}%')# 简单的推理示例print("\n--- 推理示例 ---")print(f"真实标签:{y_test[:10].cpu().numpy()}")print(f"预测标签:{predicted[:10].cpu().numpy()}")
@浙大疏锦行