liulujian/deep-risk

Fork 0

zerosaturation 4f1ff76cd1 feat: 风控程序 v1版本

2025-12-14 20:08:27 +08:00

12 KiB

Raw Blame History

收入完整性检测算法测试用例

📋 概述

本测试套件为收入完整性检测算法（RevenueIntegrityAlgorithm）提供全面的测试覆盖，包括正常场景、异常情况、边界值等共计19个测试场景。

🏗️ 测试架构

app/tests/
├── conftest.py                      # pytest配置文件，共享fixtures
└── test_revenue_integrity.py        # 主测试文件

run_tests.py                         # 测试运行脚本

📊 测试场景覆盖

1. 正常场景测试（3个用例）

✅ 收入一致（有分成协议）：充值45000元，申报45000元，预期无风险
✅ 收入一致（无分成协议）：无协议但收入匹配，预期无风险
✅ 轻微差异（低风险）：差异在可接受范围内

2. 少报收入场景测试（3个用例）

⚠️ 中度少报：差异率28.57%，风险等级MEDIUM
🚨 高度少报：差异率42.86%，风险等级HIGH
🔴 严重少报：差异率71.43%，风险等级CRITICAL

3. 多报收入场景测试（1个用例）

⚠️ 轻微多报：申报超过充值，但风险较低

4. 边界值测试（7个用例）

📌 无充值数据：有申报无充值，高风险
📌 无申报数据：有充值无申报，Critical风险
📌 完全一致：零差异，无风险
📌 大额边界值：差异100000元（刚好达到Critical阈值）
📌 差异率边界值：差异率50%（刚好达到Critical阈值）
📌 月度对账：模拟真实30笔充值场景
📌 企业主播：统一社会信用代码场景

5. 错误处理测试（4个用例）

❌ 缺少主播ID：返回UNKNOWN风险等级
❌ 缺少期间参数：参数验证错误
❌ 无效期间格式：格式校验错误
❌ 主播不存在：数据不存在错误

6. 真实业务场景测试（3个用例）

🏢 新主播无历史：少量数据处理
🏢 企业主播：企业主体测试
🏢 月度对账：大量数据场景

🚀 快速开始

1. 安装依赖

# 安装pytest和异步测试支持
pip install pytest pytest-asyncio pytest-html pytest-cov

# 安装项目依赖
pip install -r requirements.txt

2. 运行所有测试

# 使用运行脚本（推荐）
python run_tests.py

# 或直接使用pytest
pytest app/tests/test_revenue_integrity.py -v

3. 运行特定测试

# 运行正常场景测试
python run_tests.py -k "normal"

# 运行少报收入测试
python run_tests.py -k "under_reporting"

# 运行边界值测试
python run_tests.py -k "boundary"

# 运行错误处理测试
python run_tests.py -k "error"

4. 生成详细报告

# 生成HTML报告
python run_tests.py --html=test_report.html

# 生成覆盖率报告
python run_tests.py --cov

# 生成HTML覆盖率报告
python run_tests.py --cov --html=coverage_report.html

📝 测试数据说明

Mock数据结构

1. 主播信息（StreamerInfo）

{
    "streamer_id": "ZB_TEST_001",
    "streamer_name": "测试主播",
    "entity_type": "individual",  # individual/enterprise
    "tax_registration_no": "TAX123456789",
    "unified_social_credit_code": None,
    "id_card_no": "110101199001011234",
    "phone_number": "13800138000",
    "bank_account_no": "6222021234567890123",
    "bank_name": "中国工商银行",
}

2. 充值数据（PlatformRecharge）

{
    "recharge_id": "RC001",
    "user_name": "测试用户",
    "amount": 10000.0,
    "time": datetime(2024, 1, 15, 10, 30, 0),
    "payment_method": "bank_transfer",
    "status": "success",
}

3. 税务申报（TaxDeclaration）

{
    "declaration_id": "TAX001",
    "taxpayer_name": "测试主播",
    "tax_period": "2024-01",
    "sales_revenue": 45000.0,
    "declaration_date": date(2024, 2, 15),
    "tax_rate": 0.03,
}

4. 分成协议（RevenueSharingContract）

{
    "streamer_ratio": 70.0,    # 主播分成70%
    "platform_ratio": 30.0,    # 平台分成30%
    "contract_start_date": date(2024, 1, 1),
    "contract_end_date": date(2024, 12, 31),
}

🎯 风险等级判断规则

风险等级定义

风险等级	差异率阈值	差异金额阈值	描述
CRITICAL	> 50%	> 100,000元	严重风险，可能存在重大税务问题
HIGH	> 30%	> 50,000元	高度风险，需要重点关注
MEDIUM	> 10%	> 10,000元	中度风险，需要进一步核查
LOW	> 5%	> 5,000元	低度风险，持续监控
NONE	≤ 5%	≤ 5,000元	正常，收入基本一致

计算公式

# 预期收入计算（如果有分成协议）
expected_revenue = recharge_total * (streamer_ratio / 100)

# 差异计算
difference = abs(declared_revenue - expected_revenue)

# 差异率计算
difference_rate = (difference / expected_revenue) * 100

🧪 测试用例详解

示例1：测试严重少报收入

@pytest.mark.asyncio
async def test_critical_under_reporting(self, algorithm, mock_db_session, streamer_info, mock_contract_data):
    """测试场景6：严重少报收入（Critical风险）"""
    # 充值500000元，分成70%，预期350000元，申报100000元
    # 差异250000元，差异率71.43%

    recharge_data = [
        {"recharge_id": f"RC{i}", "user_name": "测试", "amount": 100000.0,
         "time": datetime(2024, 1, 15), "payment_method": "bank_transfer"}
        for i in range(1, 6)
    ]

    declaration_data = [
        {"declaration_id": "TAX001", "taxpayer_name": "测试主播",
         "tax_period": "2024-01", "sales_revenue": 100000.0,
         "declaration_date": date(2024, 2, 15)}
    ]

    mock_db_session.execute.side_effect = [
        self._create_streamer_result(streamer_info),
        self._create_recharge_result(recharge_data),
        self._create_declaration_result(declaration_data),
        self._create_contract_result(mock_contract_data),
    ]

    # 执行检测
    context = self._create_context("ZB_TEST_001", "2024-01", mock_db_session)
    result = await algorithm.detect(context)

    # 断言
    assert result.risk_level == RiskLevel.CRITICAL
    assert result.risk_score >= 85.0
    assert "严重风险" in result.description
    assert result.risk_data["difference"] >= 100000.0

示例2：测试边界值

@pytest.mark.asyncio
async def test_large_amount_boundary(self, algorithm, mock_db_session, streamer_info):
    """测试场景11：大额边界值（刚好10万元差异）"""
    # 充值200000元，申报100000元，差异100000元，差异率50%
    # 这个值刚好达到CRITICAL的金额阈值

    recharge_data = [
        {"recharge_id": "RC001", "user_name": "测试", "amount": 200000.0,
         "time": datetime(2024, 1, 15), "payment_method": "bank_transfer"}
    ]

    declaration_data = [
        {"declaration_id": "TAX001", "taxpayer_name": "测试主播",
         "tax_period": "2024-01", "sales_revenue": 100000.0,
         "declaration_date": date(2024, 2, 15)}
    ]

    mock_db_session.execute.side_effect = [
        self._create_streamer_result(streamer_info),
        self._create_recharge_result(recharge_data),
        self._create_declaration_result(declaration_data),
        MagicMock(scalar_one_or_none=lambda: None),
    ]

    # 执行检测
    context = self._create_context("ZB_TEST_001", "2024-01", mock_db_session)
    result = await algorithm.detect(context)

    # 断言：刚好达到Critical阈值
    assert result.risk_level == RiskLevel.CRITICAL
    assert result.risk_data["difference"] == 100000.0

🔧 扩展测试

添加新的测试场景

在test_revenue_integrity.py中添加新方法：

@pytest.mark.asyncio
async def test_new_scenario(self, algorithm, mock_db_session, streamer_info):
    """测试新场景：描述"""
    # 准备测试数据
    recharge_data = [...]
    declaration_data = [...]

    # Mock数据库查询
    mock_db_session.execute.side_effect = [...]

    # 执行检测
    context = self._create_context("ZB_TEST_001", "2024-01", mock_db_session)
    result = await algorithm.detect(context)

    # 断言结果
    assert result.risk_level == RiskLevel.XXX
    assert ...

在conftest.py中添加新的Mock数据：

@pytest.fixture
def mock_new_data():
    """新类型Mock数据"""
    return {
        # 数据结构定义
    }

📈 测试报告

运行测试后的输出示例

===========================================
收入完整性检测算法测试运行器
===========================================

执行命令: python -m pytest app/tests/test_revenue_integrity.py -v

collected 19 items

app/tests/test_revenue_integrity.py::TestRevenueIntegrityNormal::test_revenue_match_with_contract PASSED [  5%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityNormal::test_revenue_match_without_contract PASSED [ 10%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityNormal::test_slight_difference_low_risk PASSED [ 15%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityUnderReporting::test_medium_under_reporting PASSED [ 21%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityUnderReporting::test_high_under_reporting PASSED [ 26%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityUnderReporting::test_critical_under_reporting PASSED [ 31%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityOverReporting::test_over_reporting_low_risk PASSED [ 36%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityEdgeCases::test_no_recharge_data PASSED [ 42%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityEdgeCases::test_no_declaration_data PASSED [ 47%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityEdgeCases::test_zero_difference PASSED [ 52%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityEdgeCases::test_large_amount_boundary PASSED [ 57%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityEdgeCases::test_rate_boundary_50_percent PASSED [ 63%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityErrorHandling::test_missing_streamer_id PASSED [ 68%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityErrorHandling::test_missing_period PASSED [ 73%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityErrorHandling::test_invalid_period_format PASSED [ 78%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityErrorHandling::test_streamer_not_found PASSED [ 84%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityBusinessScenarios::test_monthly_reconciliation PASSED [ 89%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityBusinessScenarios::test_new_streamer_no_history PASSED [ 94%]
app/tests/test_revenue_integrity.py::TestRevenueIntegrityBusinessScenarios::test_enterprise_streamer PASSED [100%]

===========================================
✓ 所有测试通过！
===========================================

⚠️ 注意事项

异步测试：所有测试都是异步的，需要使用@pytest.mark.asyncio标记
数据库Mock：使用AsyncMock模拟数据库会话，避免真实数据库连接
数据隔离：每次测试使用独立的Mock数据，避免数据污染
风险等级断言：确保断言的风险等级与测试场景匹配
边界值测试：特别注意差异率和金额的边界值

🐛 常见问题

Q: 测试运行失败，提示"ModuleNotFoundError"

A: 确保在项目根目录运行，并检查PYTHONPATH设置

export PYTHONPATH=/path/to/deeprisk-claude-1/backend:$PYTHONPATH

Q: 异步测试报错

A: 确保安装了pytest-asyncio并使用正确的标记

pip install pytest-asyncio

Q: 想要调试特定测试

A: 使用pytest的调试选项

pytest app/tests/test_revenue_integrity.py::TestRevenueIntegrityNormal::test_revenue_match_with_contract -v -s

📚 参考资料

🤝 贡献

欢迎提交新的测试用例！请确保：

测试场景明确
Mock数据完整
断言准确
遵循现有代码风格

创建时间: 2024-11-30 版本: 1.0.0 作者: Claude Code

12 KiB Raw Blame History Unescape Escape