根据人行征信,做特征衍生,写3个指标的中英文名,并且写出指标的特征类型,以及特征的风险类型,对每个特效写伪代码指标逻辑和sql逻辑,特征衍生的经典一点,独特一点
时间: 2024-06-12 21:08:23 浏览: 18
指标一:负债比例(Debt Ratio)
特征类型:数值型
风险类型:风险越高,负债比例越高
伪代码指标逻辑:
- 计算个人所有贷款的总额
- 计算个人所有信用卡的总额
- 计算个人所有负债的总额
- 计算个人所有资产的总额
- 计算负债比例 = 总负债 / 总资产
SQL逻辑:
```
SELECT SUM(loan_amount) AS total_loan, SUM(credit_card_amount) AS total_credit_card,
SUM(loan_amount + credit_card_amount) AS total_debt, SUM(asset_amount) AS total_asset,
SUM(loan_amount + credit_card_amount) / SUM(asset_amount) AS debt_ratio
FROM personal_credit_info
WHERE user_id = 'xxx';
```
指标二:信用卡使用率(Credit Utilization Ratio)
特征类型:数值型
风险类型:风险越高,信用卡使用率越高
伪代码指标逻辑:
- 计算个人信用卡已使用额度
- 计算个人信用卡总额度
- 计算信用卡使用率 = 已使用额度 / 总额度
SQL逻辑:
```
SELECT SUM(used_credit) AS total_used_credit, SUM(total_credit) AS total_credit,
SUM(used_credit) / SUM(total_credit) AS credit_utilization_ratio
FROM personal_credit_info
WHERE user_id = 'xxx' AND product_type = 'credit_card';
```
指标三:逾期次数(Overdue Times)
特征类型:离散型
风险类型:风险越高,逾期次数越多
伪代码指标逻辑:
- 统计个人所有贷款和信用卡的逾期次数
SQL逻辑:
```
SELECT COUNT(*) AS overdue_times
FROM personal_credit_info
WHERE user_id = 'xxx' AND overdue_days > 0;
```