From 88186f37a60dda07eec317cdcbfc3e75843b88f5 Mon Sep 17 00:00:00 2001
From: wkc <978997012@qq.com>
Date: Tue, 17 Mar 2026 16:59:39 +0800
Subject: [PATCH] =?UTF-8?q?=E6=8F=90=E4=BA=A4=E6=A8=A1=E5=9E=8BSQL?=
 =?UTF-8?q?=E6=A0=B8=E5=AF=B9=E4=B8=8E=E6=94=B9=E5=86=99=E6=8A=A5=E5=91=8A?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../2026-03-17-model-sql-check-and-rewrite.md | 1260 +++++++++++++++++
 1 file changed, 1260 insertions(+)
 create mode 100644 docs/reports/implementation/2026-03-17-model-sql-check-and-rewrite.md

diff --git a/docs/reports/implementation/2026-03-17-model-sql-check-and-rewrite.md b/docs/reports/implementation/2026-03-17-model-sql-check-and-rewrite.md
new file mode 100644
index 00000000..b2b04f45
--- /dev/null
+++ b/docs/reports/implementation/2026-03-17-model-sql-check-and-rewrite.md
@@ -0,0 +1,1260 @@
+# 模型 SQL 与当前库表结构核对及改写报告
+
+## 执行概况
+
+- 执行时间：2026-03-17
+- 源文件：[assets/模型信息.csv](/Users/wkc/Desktop/ccdi/ccdi/assets/模型信息.csv)
+- 数据库配置来源：[ruoyi-admin/src/main/resources/application-dev.yml](/Users/wkc/Desktop/ccdi/ccdi/ruoyi-admin/src/main/resources/application-dev.yml)
+- 目标数据库：`MySQL 5.7.44`，库名 `ccdi`
+- 报告目标：
+  - 保留 CSV 原始内容并转存为 Markdown
+  - 检查模型 SQL 与当前数据库表结构是否匹配
+  - 检查 SQL 在当前 MySQL 方言下的语法兼容性
+  - 将可落地 SQL 改写为适配当前库结构的版本
+
+## 核对基线
+
+### 当前库存在的核心表
+
+- `ccdi_base_staff`
+- `ccdi_bank_statement`
+- `ccdi_staff_fmy_relation`
+- `ccdi_asset_info`
+- `ccdi_biz_intermediary`
+- `ccdi_cust_fmy_relation`
+- `ccdi_cust_enterprise_relation`
+
+### 当前库不存在的外部对象
+
+- `odsdb.blfmconf`
+- `sjfx_pro.bdfmhqaa_orc`
+- `xdzx.assure_infomation`
+- `中介名单`
+
+### 当前库关键字段实际情况
+
+- `ccdi_bank_statement` 存在 `TRX_DATE`，不存在 `trx_time`
+- `ccdi_asset_info` 存在 `update_time`，不存在 `updated_at`
+- `ccdi_bank_statement` 的金额字段为 `AMOUNT_DR`、`AMOUNT_CR`
+- `ccdi_bank_statement` 的摘要字段为 `USER_MEMO`
+- `ccdi_bank_statement` 的对手方字段为 `CUSTOMER_ACCOUNT_NAME`、`CUSTOMER_ACCOUNT_NO`
+
+### 本次改写原则
+
+- 保持原始业务意图，不随意增加新的业务规则
+- 优先适配 `MySQL 5.7`，去除 `WITH`、`add_months()` 等不兼容写法
+- 将 `PROJECT_ID`、`$$$`、`$$$$` 改为 MySQL 会话变量
+- 对于当前库无法独立支持的外部依赖，给出“部分可执行”的改写版，并明确缺失项
+- 对于原 SQL 与业务口径不完全一致的地方，在说明中单独标注
+
+---
+
+## CSV 原始内容与逐条改写
+
+### CSV 第 1 行
+
+- 序号：`2.1`
+- 模型名称：`异常交易`
+- 核心异常点（展示在前端页面）：`与客户之间非正常资金往来`
+- 业务口径：`员工及关系人与客户及关系人之间有超过1000元以上的资金往来；客户指信贷类客户包括贷款户、担保人，中介库人员，包括中介注册的主体及主体关系人。`
+- 相关指标：`与特定客户交易总额`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人及亲属、贷款客户、担保人、中介`
+- 技术口径：`空`
+- 限制阈值指标：`/`
+- 可疑结果返回：`流水明细`
+- 风险等级：`高风险`
+- 原始代码：
+
+```sql
+--员工及其亲属与贷款客户、担保户、中介有异常交易
+with loan_cust_acct as (
+select t2.aa01ac15
+from
+(
+select substr(nfabcsid,4) as nfabcsid
+    from odsdb.blfmconf --贷款合同文件
+    where nfaacost in ('3','5','7') --合同状态
+    and substr(nfaabrno,1,3) = '902' --机构
+    and del_f = '0'
+    group by substr(nfabcsid,4)
+) t1
+inner join
+(
+select
+        aa01ac15        --账号
+,aa62cfno
+from sjfx_pro.bdfmhqaa_orc
+where del_f = 0
+    and substr(trim(aa47brno),1,3) = '902'     --机构号修改
+and rcstrs1b <>'9'
+    AND aa15zhzt ='1'    -- 账户状态 1-正常 2-销户 3-新开户 4-结清
+group by  aa01ac15 ,aa62cfno
+) t2
+on t1.nfabcsid = t2.aa62cfno
+) ,
+assure_cust_acct as (
+select t2.aa01ac15
+from
+(
+select asseure_sign
+from xdzx.assure_infomation
+where del_f= '0'
+and assure_state <> '2'
+and substr(create_org,1,3)='902'
+group by asseure_sign
+) t1
+inner join
+(
+select
+        aa01ac15        --账号
+,aa03csno
+from sjfx_pro.bdfmhqaa_orc
+where del_f = 0
+    and substr(trim(aa47brno),1,3) = '902'     --机构号修改
+and rcstrs1b <>'9'
+    AND aa15zhzt ='1'    -- 账户状态 1-正常 2-销户 3-新开户 4-结清
+group by  aa01ac15 ,aa03csno
+) t2
+on t1.asseure_sign = t2.aa03csno
+)
+select a.bank_statement_id
+from
+(
+select t1.id_card
+,t2.bank_statement_id
+,customer_account_no
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where project_id = PROJECT_ID
+and (amount_dr>1000 or amount_cr>1000)
+and customer_account_no is not null
+union all
+select t1.person_id
+,t2.bank_statement_id
+,customer_account_no
+from
+ccdi_staff_fmy_relation t1
+inner join
+ccdi_bank_statement t2
+on t1.relation_cert_no = t2.cret_no
+where t1.status = 1
+and t2.project_id = PROJECT_ID
+and (amount_dr>1000 or amount_cr>1000)
+and customer_account_no is not null
+) a
+where exists (select 1 from loan_cust_acct b on a.customer_account_no = b.aa01ac15)
+or exists (select 1 from assure_cust_acct c on a.customer_account_no = c.aa01ac15)
+or exists (select 1 from 中介名单 d on a.customer_account_no = d.中介账号);
+```
+
+**核对结论**
+
+- 原 SQL 依赖 3 个当前库不存在的外部 schema 和 1 个不存在的中介表
+- 原 SQL 使用 `WITH`，当前 `MySQL 5.7` 不支持
+- 原 SQL 中 `exists (select 1 from ... on ...)` 是明确语法错误
+- 当前 `ccdi` 库没有贷款客户账号、担保人账号、中介账号的现成对照表，无法完全在现库独立落地
+
+**改写 SQL**
+
+以下改写版可在 `MySQL 5.7` 执行，但需要先准备 3 张引用表：
+
+- `tmp_model_21_loan_cust_acct(account_no)`
+- `tmp_model_21_assure_cust_acct(account_no)`
+- `tmp_model_21_intermediary_acct(account_no)`
+
+```sql
+SET @project_id := 0;
+SET @min_trade_amount := 1000;
+
+SELECT DISTINCT a.bank_statement_id
+FROM (
+    SELECT
+        t1.id_card AS person_id,
+        t2.bank_statement_id,
+        t2.CUSTOMER_ACCOUNT_NO AS customer_account_no
+    FROM ccdi_base_staff t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.id_card = t2.cret_no
+    WHERE t2.project_id = @project_id
+      AND (t2.AMOUNT_DR > @min_trade_amount OR t2.AMOUNT_CR > @min_trade_amount)
+      AND t2.CUSTOMER_ACCOUNT_NO IS NOT NULL
+
+    UNION ALL
+
+    SELECT
+        t1.person_id,
+        t2.bank_statement_id,
+        t2.CUSTOMER_ACCOUNT_NO AS customer_account_no
+    FROM ccdi_staff_fmy_relation t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.relation_cert_no = t2.cret_no
+    WHERE t1.status = 1
+      AND t2.project_id = @project_id
+      AND (t2.AMOUNT_DR > @min_trade_amount OR t2.AMOUNT_CR > @min_trade_amount)
+      AND t2.CUSTOMER_ACCOUNT_NO IS NOT NULL
+) a
+WHERE EXISTS (
+    SELECT 1
+    FROM tmp_model_21_loan_cust_acct b
+    WHERE b.account_no = a.customer_account_no
+)
+   OR EXISTS (
+    SELECT 1
+    FROM tmp_model_21_assure_cust_acct c
+    WHERE c.account_no = a.customer_account_no
+)
+   OR EXISTS (
+    SELECT 1
+    FROM tmp_model_21_intermediary_acct d
+    WHERE d.account_no = a.customer_account_no
+);
+```
+
+**未解决依赖**
+
+- 当前库没有贷款客户账号、担保人账号、中介账号映射表
+- 如果后续需要完全落地，应先把外部客户账号清单同步到 `ccdi` 或建立中间表
+
+---
+
+### CSV 第 2 行
+
+- 序号：`2.2`
+- 模型名称：`空`
+- 核心异常点（展示在前端页面）：`低收入亲属大额交易`
+- 业务口径：`关系人中没有收入或月收入低于3000元的人员，累计交易金额超过10万元。`
+- 相关指标：`低收入亲属的累计交易额`
+- 指标英文名：`空`
+- 风险筛查对象：`员工亲属`
+- 技术口径：`关联员工及其亲属，筛选平均月收入<3000的亲属的账户，在流水中单笔或累计 >10 万元，按员工汇总`
+- 限制阈值指标：`/`
+- 可疑结果返回：`个人、累积金额`
+- 风险等级：`一般`
+- 原始代码：
+
+```sql
+--员工亲属低收入但交易金额高
+select t1.person_id
+,t2.trans_amount
+from
+(
+select person_id
+,relation_cert_no
+,avg(amount_cr) as avg_amount_cr
+from
+(
+select t1.person_id
+,t1.relation_cert_no
+,substr(trx_time,1,7)
+,sum(amount_cr) as amount_cr--收入金额
+from
+ccdi_staff_fmy_relation t1
+inner join
+ccdi_bank_statement t2
+on t1.relation_cert_no = t2.cret_no
+where t1.status = 1
+and t2.project_id = PROJECT_ID
+and t2.le_account_name <> t2.customer_account_name --排除同名交易
+group by t1.person_id,t1.relation_cert_no,substr(trx_time,1,7)
+)
+group by person_id,relation_cert_no
+having avg(amount_cr)<=3000 ---月均收入不超过3000
+) t1
+left join
+(
+select t1.relation_cert_no
+,sum(amount_cr + amount_dr) as trans_amount
+,max(amount_cr) as max_amount_cr
+,max(amount_dr) as max_amount_dr
+from
+ccdi_staff_fmy_relation t1
+inner join
+ccdi_bank_statement t2
+on t1.relation_cert_no = t2.cret_no
+where t1.status = 1
+and t2.project_id = PROJECT_ID
+group by t1.relation_cert_no
+having sum(amount_cr + amount_dr) >= 100000
+) t2
+on t1.relation_cert_no = t2.relation_cert_no
+where t2.relation_cert_no is not null
+;
+```
+
+**核对结论**
+
+- 原 SQL 引用了不存在的 `trx_time`
+- 原 SQL 的内层派生表缺少别名
+- 业务意图可完全映射到当前库
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+SET @monthly_income_ceiling := 3000;
+SET @total_trade_floor := 100000;
+
+SELECT
+    low_income.person_id,
+    low_income.relation_cert_no,
+    risky.trans_amount,
+    risky.max_amount_cr,
+    risky.max_amount_dr
+FROM (
+    SELECT
+        monthly.person_id,
+        monthly.relation_cert_no,
+        AVG(monthly.month_income) AS avg_amount_cr
+    FROM (
+        SELECT
+            t1.person_id,
+            t1.relation_cert_no,
+            LEFT(t2.TRX_DATE, 7) AS income_month,
+            SUM(t2.AMOUNT_CR) AS month_income
+        FROM ccdi_staff_fmy_relation t1
+        INNER JOIN ccdi_bank_statement t2
+            ON t1.relation_cert_no = t2.cret_no
+        WHERE t1.status = 1
+          AND t2.project_id = @project_id
+          AND t2.LE_ACCOUNT_NAME <> t2.CUSTOMER_ACCOUNT_NAME
+        GROUP BY t1.person_id, t1.relation_cert_no, LEFT(t2.TRX_DATE, 7)
+    ) monthly
+    GROUP BY monthly.person_id, monthly.relation_cert_no
+    HAVING AVG(monthly.month_income) <= @monthly_income_ceiling
+) low_income
+INNER JOIN (
+    SELECT
+        t1.relation_cert_no,
+        SUM(IFNULL(t2.AMOUNT_CR, 0) + IFNULL(t2.AMOUNT_DR, 0)) AS trans_amount,
+        MAX(t2.AMOUNT_CR) AS max_amount_cr,
+        MAX(t2.AMOUNT_DR) AS max_amount_dr
+    FROM ccdi_staff_fmy_relation t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.relation_cert_no = t2.cret_no
+    WHERE t1.status = 1
+      AND t2.project_id = @project_id
+    GROUP BY t1.relation_cert_no
+    HAVING SUM(IFNULL(t2.AMOUNT_CR, 0) + IFNULL(t2.AMOUNT_DR, 0)) >= @total_trade_floor
+) risky
+    ON low_income.relation_cert_no = risky.relation_cert_no;
+```
+
+---
+
+### CSV 第 3 行
+
+- 序号：`3.1`
+- 模型名称：`疑似赌博`
+- 核心异常点（展示在前端页面）：`疑似赌博交易`
+- 业务口径：`多人（2人及以上）、多次（2次以上）、相近时间（同一天）内，有转账、微信转账、支付宝转账发生，且额度在可疑区间。金额区间可在排查设置页面进行设置`
+- 相关指标：`单日疑似赌博转账金额下限`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人`
+- 技术口径：`员工本人账户中，24h内统计向不同对手方转账（cash_type='转账'）且金额在可疑区间，笔数≥2且对手方≥2`
+- 限制阈值指标：`单日频繁转账金额下限`
+- 可疑结果返回：`个人、日期、累计金额`
+- 风险等级：`高风险`
+- 原始代码：
+
+```sql
+--员工 疑似赌博
+select id_card,trans_date
+from
+(
+select t1.id_card,left(trx_time,10) as trans_date,customer_account_name
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where project_id = PROJECT_ID
+and ((amount_dr>= $$$ and amount_dr<=$$$) or (amount_cr>= $$$ and amount_cr<=$$$))  -----转入转出金额区间
+and customer_account_name is not null
+and (user_memo rlike '转帐|转账|红包|网转|转入' or cash_type rlike '转帐|转账|红包|网转|转入')
+and user_memo not like '%款%'
+and t2.le_account_name <> t2.customer_account_name --排除同名交易
+group by t1.id_card,left(trx_time,10),customer_account_name
+having count(1)>2
+)
+group by id_card,trans_date
+having count(distinct customer_account_name)>=2;
+```
+
+**核对结论**
+
+- 原 SQL 引用了不存在的 `trx_time`
+- 外层 `from (...)` 缺少别名
+- 原技术口径写“24h 内”，原 SQL 实际按自然日统计
+- 原 SQL 内层 `having count(1)>2` 与“2次以上”的自然语言描述并不完全一致
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+SET @amount_min := 1000;
+SET @amount_max := 5000;
+
+SELECT
+    daily.id_card,
+    daily.trans_date,
+    COUNT(DISTINCT daily.customer_account_name) AS counterparty_count,
+    SUM(daily.trade_count) AS trade_count
+FROM (
+    SELECT
+        t1.id_card,
+        LEFT(t2.TRX_DATE, 10) AS trans_date,
+        t2.CUSTOMER_ACCOUNT_NAME AS customer_account_name,
+        COUNT(*) AS trade_count
+    FROM ccdi_base_staff t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.id_card = t2.cret_no
+    WHERE t2.project_id = @project_id
+      AND (
+            (t2.AMOUNT_DR BETWEEN @amount_min AND @amount_max)
+         OR (t2.AMOUNT_CR BETWEEN @amount_min AND @amount_max)
+      )
+      AND t2.CUSTOMER_ACCOUNT_NAME IS NOT NULL
+      AND (
+            t2.USER_MEMO REGEXP '转帐|转账|红包|网转|转入'
+         OR t2.CASH_TYPE REGEXP '转帐|转账|红包|网转|转入'
+      )
+      AND (t2.USER_MEMO IS NULL OR t2.USER_MEMO NOT LIKE '%款%')
+      AND t2.LE_ACCOUNT_NAME <> t2.CUSTOMER_ACCOUNT_NAME
+    GROUP BY t1.id_card, LEFT(t2.TRX_DATE, 10), t2.CUSTOMER_ACCOUNT_NAME
+    HAVING COUNT(*) >= 2
+) daily
+GROUP BY daily.id_card, daily.trans_date
+HAVING COUNT(DISTINCT daily.customer_account_name) >= 2;
+```
+
+**说明**
+
+- 改写版将“多次”按 `>= 2` 处理，更贴近中文业务口径
+- 如果后续必须按原 SQL 的严格条件执行，可将 `HAVING COUNT(*) >= 2` 改回 `> 2`
+
+---
+
+### CSV 第 4 行
+
+- 序号：`空`
+- 模型名称：`空`
+- 核心异常点（展示在前端页面）：`空`
+- 业务口径：`空`
+- 相关指标：`单日疑似赌博转账金额上限`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人`
+- 技术口径：`空`
+- 代码：`空`
+- 限制阈值指标：`单日频繁转账金额上限`
+- 可疑结果返回：`空`
+- 风险等级：`空`
+
+**核对结论**
+
+- 该行不是独立模型 SQL，而是对 `3.1` 的上限参数补充
+- 建议在模型参数表中为 `3.1` 同时维护下限和上限两个阈值字段，不应单独落成 SQL
+
+---
+
+### CSV 第 5 行
+
+- 序号：`3.2`
+- 模型名称：`空`
+- 核心异常点（展示在前端页面）：`疑似敏感交易`
+- 业务口径：`备注或交易摘要、对手有“游戏、抖币、体彩、福彩”等字眼。`
+- 相关指标：`赌博关键词支出`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人`
+- 技术口径：`员工本人账户中，筛选备注或交易摘要、对手方名称有“游戏、抖币、体彩、福彩”等字眼关键词，汇总金额`
+- 限制阈值指标：`/`
+- 可疑结果返回：`流水明细`
+- 风险等级：`高风险`
+- 原始代码：
+
+```sql
+--员工 网络赌博、体彩
+select t2.bank_statement_id
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where project_id = PROJECT_ID
+and amount_dr> 0
+and (user_memo rlike '游戏|抖币|体彩|福彩|彩票|赌|球|外围|博彩|六合|时时彩|赛车|赌场|筹码|盘口|返水|洗码|庄家|闲家|百家乐|斗牛|炸金花|牌九|麻将|捕鱼|电子游艺|投注'
+or customer_account_name rlike '游戏|抖币|体彩|福彩|彩票|赌|球|外围|博彩|六合|时时彩|赛车|赌场|筹码|盘口|返水|洗码|庄家|闲家|百家乐|斗牛|炸金花|牌九|麻将|捕鱼|电子游艺|投注')
+;
+```
+
+**核对结论**
+
+- 主表和字段与当前库匹配
+- 语法可兼容 `MySQL 5.7`
+- 仍需用参数替换 `PROJECT_ID`
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+
+SELECT
+    t2.bank_statement_id
+FROM ccdi_base_staff t1
+INNER JOIN ccdi_bank_statement t2
+    ON t1.id_card = t2.cret_no
+WHERE t2.project_id = @project_id
+  AND t2.AMOUNT_DR > 0
+  AND (
+        t2.USER_MEMO REGEXP '游戏|抖币|体彩|福彩|彩票|赌|球|外围|博彩|六合|时时彩|赛车|赌场|筹码|盘口|返水|洗码|庄家|闲家|百家乐|斗牛|炸金花|牌九|麻将|捕鱼|电子游艺|投注'
+     OR t2.CUSTOMER_ACCOUNT_NAME REGEXP '游戏|抖币|体彩|福彩|彩票|赌|球|外围|博彩|六合|时时彩|赛车|赌场|筹码|盘口|返水|洗码|庄家|闲家|百家乐|斗牛|炸金花|牌九|麻将|捕鱼|电子游艺|投注'
+  );
+```
+
+---
+
+### CSV 第 6 行
+
+- 序号：`4`
+- 模型名称：`可疑关系`
+- 核心异常点（展示在前端页面）：`特殊金额交易`
+- 业务口径：`除与配偶、子女外，发生特殊金额交易，如1314元、520元等具有特殊含义的金额。`
+- 相关指标：`特殊金额交易`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人`
+- 技术口径：`员工本人账户中，筛选交易金额匹配特殊值（如1314），且对手方名称非配偶/子女，统计笔数`
+- 限制阈值指标：`/`
+- 可疑结果返回：`流水明细`
+- 风险等级：`空`
+- 原始代码：
+
+```sql
+--员工 可疑关系
+select distinct t2.bank_statement_id
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+left join
+ccdi_staff_fmy_relation t3
+on t1.id_card = t3.person_id
+and t2.customer_account_name = t3.relation_name
+where t2.project_id = PROJECT_ID
+and t2.le_account_name <> t2.customer_account_name --排除同名交易
+and t3.relation_type not in (配偶，子女) --关系类型按实际数据的码值确定
+and (amount_dr in (520,1314) or amount_cr in (520,1314))
+;
+```
+
+**核对结论**
+
+- 原 SQL 中 `not in (配偶，子女)` 缺少引号，且使用了全角逗号，语法错误
+- 使用 `left join + not in` 会把未匹配到亲属关系的记录误伤为 `null not in (...)`
+- 当前库的 `relation_type` 实际可能存码值，不能直接假定一定是中文
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+
+SELECT DISTINCT
+    t2.bank_statement_id
+FROM ccdi_base_staff t1
+INNER JOIN ccdi_bank_statement t2
+    ON t1.id_card = t2.cret_no
+WHERE t2.project_id = @project_id
+  AND t2.LE_ACCOUNT_NAME <> t2.CUSTOMER_ACCOUNT_NAME
+  AND (t2.AMOUNT_DR IN (520, 1314) OR t2.AMOUNT_CR IN (520, 1314))
+  AND NOT EXISTS (
+      SELECT 1
+      FROM ccdi_staff_fmy_relation t3
+      WHERE t3.person_id = t1.id_card
+        AND t3.status = 1
+        AND t3.relation_name = t2.CUSTOMER_ACCOUNT_NAME
+        AND t3.relation_type IN ('配偶', '子女')
+  );
+```
+
+**说明**
+
+- 如果 `relation_type` 在现网存的是码值，需要把 `'配偶'`、`'子女'` 替换成真实码值
+
+---
+
+### CSV 第 7 行
+
+- 序号：`5.1`
+- 模型名称：`可疑兼职`
+- 核心异常点（展示在前端页面）：`疑似兼职`
+- 业务口径：`除本行工资收入外，每月有固定收入，固定收入金额自行设置。`
+- 相关指标：`月度非本行工资收入金额`
+- 指标英文名：`MONTHLY_FIXED_INCOME`
+- 风险筛查对象：`员工本人`
+- 技术口径：`员工本人账户中，筛选 amount_cr>0，除本行工资收入外，每月有固定收入，固定收入金额自行设置。`
+- 限制阈值指标：`非本行工资收入金额`
+- 可疑结果返回：`个人、累积金额`
+- 风险等级：`空`
+- 原始代码：
+
+```sql
+--员工 可疑兼职
+select id_card
+FROM
+(
+select t1.id_card
+,left(TRX_DATE,7) as incm_mnth
+,sum(t2.amount_cr) as sum_amount_cr
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where project_id = PROJECT_ID
+and t2.le_account_name <> t2.customer_account_name --排除同名交易
+and not (customer_account_name = '浙江兰溪农村商业银行股份有限公司'
+and (user_memo rlike '代发|工资|分红|红利|奖金|薪酬|薪金|补贴|薪|年终奖|年金|加班费|劳务费|劳务外包|提成|劳务派遣|绩效|酬劳|批量代付|PAYROLL|SALA|CPF|directors.*fees'
+or cash_type rlike '代发|工资|劳务费' ))
+and customer_account_name is not null
+and left(TRX_DATE,10) >= add_months(current_date(),-12)     --近一年
+and amount_cr>0
+group by t1.id_card,left(t2.TRX_DATE,7)
+)
+where sum_amount_cr > $$$$ --每月收入参数
+group by id_card
+having count(distinct incm_mnth)>=6   -----一年内6个月满足收入门槛
+and STDDEV(sum_amount_cr)/AVG(sum_amount_cr) <= 0.3  --标准差与平均值的比值小，波动小
+;
+```
+
+**核对结论**
+
+- 原 SQL 外层派生表缺少别名
+- 原 SQL 使用 `add_months()`，当前 `MySQL 5.7` 不支持
+- `TRX_DATE` 为字符串，必须显式转日期后再做时间比较
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+SET @monthly_income_floor := 5000;
+SET @stable_month_count := 6;
+SET @volatility_ratio_limit := 0.3;
+
+SELECT
+    monthly_income.id_card
+FROM (
+    SELECT
+        t1.id_card,
+        LEFT(t2.TRX_DATE, 7) AS incm_mnth,
+        SUM(t2.AMOUNT_CR) AS sum_amount_cr
+    FROM ccdi_base_staff t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.id_card = t2.cret_no
+    WHERE t2.project_id = @project_id
+      AND t2.LE_ACCOUNT_NAME <> t2.CUSTOMER_ACCOUNT_NAME
+      AND NOT (
+            t2.CUSTOMER_ACCOUNT_NAME = '浙江兰溪农村商业银行股份有限公司'
+        AND (
+                t2.USER_MEMO REGEXP '代发|工资|分红|红利|奖金|薪酬|薪金|补贴|薪|年终奖|年金|加班费|劳务费|劳务外包|提成|劳务派遣|绩效|酬劳|批量代付|PAYROLL|SALA|CPF|directors.*fees'
+             OR t2.CASH_TYPE REGEXP '代发|工资|劳务费'
+        )
+      )
+      AND t2.CUSTOMER_ACCOUNT_NAME IS NOT NULL
+      AND STR_TO_DATE(LEFT(t2.TRX_DATE, 10), '%Y-%m-%d') >= DATE_SUB(CURDATE(), INTERVAL 12 MONTH)
+      AND t2.AMOUNT_CR > 0
+    GROUP BY t1.id_card, LEFT(t2.TRX_DATE, 7)
+) monthly_income
+WHERE monthly_income.sum_amount_cr > @monthly_income_floor
+GROUP BY monthly_income.id_card
+HAVING COUNT(DISTINCT monthly_income.incm_mnth) >= @stable_month_count
+   AND STDDEV(monthly_income.sum_amount_cr) / NULLIF(AVG(monthly_income.sum_amount_cr), 0) <= @volatility_ratio_limit;
+```
+
+---
+
+### CSV 第 8 行
+
+- 序号：`5.2`
+- 模型名称：`空`
+- 核心异常点（展示在前端页面）：`疑似兼职`
+- 业务口径：`每季或每年从固定交易对手转入金额，金额可设区间值，如5000-10000。`
+- 相关指标：`季度稳定收入金额`
+- 指标英文名：`FIXED_COUNTERPARTY_TRANSFER`
+- 风险筛查对象：`员工本人`
+- 技术口径：`员工本人账户中，按季/年统计同一对手方累计流入金额在设定区间内`
+- 限制阈值指标：`稳定季度收入金额`
+- 可疑结果返回：`个人、累积金额`
+- 风险等级：`空`
+- 原始代码：
+
+```sql
+--员工 可疑固定收入
+select id_card
+from
+(
+select
+id_card
+,customer_account_name
+,trans_qrt
+,sum(amount_cr) as amount_total
+from
+(
+select t1.id_card
+,customer_account_name
+,amount_cr
+,concat(year(trx_time),'-Q',quarter(trx_time)) as trans_qrt  --每季度的固定收入
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where project_id = PROJECT_ID
+and not (customer_account_name = '浙江兰溪农村商业银行股份有限公司'
+and (user_memo rlike '代发|工资|分红|红利|奖金|薪酬|薪金|补贴|薪|年终奖|年金|加班费|劳务费|劳务外包|提成|劳务派遣|绩效|酬劳|批量代付|PAYROLL|SALA|CPF|directors.*fees'
+or cash_type rlike '代发|工资|劳务费' ))
+and customer_account_name is not null
+and left(TRX_DATE,10) >= add_months(current_date(),-12)     --近一年
+and amount_cr>0   --每笔收入金额
+and t2.le_account_name <> t2.customer_account_name --排除同名交易
+)
+group by id_card,customer_account_name,trans_qrt
+having sum(amount_cr) between $$$$ and $$$$    ---该季度总收入，区间自行设置
+)
+group by id_card
+having count(distinct customer_account_name) < 3
+;
+```
+
+**核对结论**
+
+- 原 SQL 引用了不存在的 `trx_time`
+- 两层派生表均缺少别名
+- 原 SQL 使用 `add_months()`，当前 `MySQL 5.7` 不支持
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+SET @quarter_income_min := 5000;
+SET @quarter_income_max := 10000;
+
+SELECT
+    quarter_income.id_card
+FROM (
+    SELECT
+        src.id_card,
+        src.customer_account_name,
+        CONCAT(YEAR(src.trx_date), '-Q', QUARTER(src.trx_date)) AS trans_qrt,
+        SUM(src.amount_cr) AS amount_total
+    FROM (
+        SELECT
+            t1.id_card,
+            t2.CUSTOMER_ACCOUNT_NAME AS customer_account_name,
+            t2.AMOUNT_CR AS amount_cr,
+            STR_TO_DATE(LEFT(t2.TRX_DATE, 10), '%Y-%m-%d') AS trx_date
+        FROM ccdi_base_staff t1
+        INNER JOIN ccdi_bank_statement t2
+            ON t1.id_card = t2.cret_no
+        WHERE t2.project_id = @project_id
+          AND NOT (
+                t2.CUSTOMER_ACCOUNT_NAME = '浙江兰溪农村商业银行股份有限公司'
+            AND (
+                    t2.USER_MEMO REGEXP '代发|工资|分红|红利|奖金|薪酬|薪金|补贴|薪|年终奖|年金|加班费|劳务费|劳务外包|提成|劳务派遣|绩效|酬劳|批量代付|PAYROLL|SALA|CPF|directors.*fees'
+                 OR t2.CASH_TYPE REGEXP '代发|工资|劳务费'
+            )
+          )
+          AND t2.CUSTOMER_ACCOUNT_NAME IS NOT NULL
+          AND STR_TO_DATE(LEFT(t2.TRX_DATE, 10), '%Y-%m-%d') >= DATE_SUB(CURDATE(), INTERVAL 12 MONTH)
+          AND t2.AMOUNT_CR > 0
+          AND t2.LE_ACCOUNT_NAME <> t2.CUSTOMER_ACCOUNT_NAME
+    ) src
+    GROUP BY src.id_card, src.customer_account_name, CONCAT(YEAR(src.trx_date), '-Q', QUARTER(src.trx_date))
+    HAVING SUM(src.amount_cr) BETWEEN @quarter_income_min AND @quarter_income_max
+) quarter_income
+GROUP BY quarter_income.id_card
+HAVING COUNT(DISTINCT quarter_income.customer_account_name) < 3;
+```
+
+---
+
+### CSV 第 9 行
+
+- 序号：`空`
+- 模型名称：`空`
+- 核心异常点（展示在前端页面）：`疑似兼职`
+- 业务口径：`转入资金摘要有“工资”、“分红”、“红利”、“利息（非银行结息）”等收入`
+- 相关指标：`/`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人`
+- 技术口径：`空`
+- 限制阈值指标：`/`
+- 可疑结果返回：`流水明细`
+- 风险等级：`高风险`
+- 原始代码：
+
+```sql
+select distinct t2.bank_statement_id
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where t2.project_id = PROJECT_ID
+and t2.customer_account_name <> '浙江兰溪农村商业银行股份有限公司'
+and (t2.user_memo rlike '代发|工资|分红|红利|奖金|薪酬|薪金|补贴|薪|年终奖|年金|加班费|劳务费|劳务外包|提成|劳务派遣|绩效|酬劳|批量代付|PAYROLL|SALA|CPF|directors.*fees'
+or cash_type rlike '代发|工资|劳务费' )
+;
+```
+
+**核对结论**
+
+- 原 SQL 主表、字段、语法均基本匹配当前库
+- 结合业务口径，建议增加 `AMOUNT_CR > 0`，避免把支出误判成收入类线索
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+
+SELECT DISTINCT
+    t2.bank_statement_id
+FROM ccdi_base_staff t1
+INNER JOIN ccdi_bank_statement t2
+    ON t1.id_card = t2.cret_no
+WHERE t2.project_id = @project_id
+  AND t2.AMOUNT_CR > 0
+  AND t2.CUSTOMER_ACCOUNT_NAME <> '浙江兰溪农村商业银行股份有限公司'
+  AND (
+        t2.USER_MEMO REGEXP '代发|工资|分红|红利|奖金|薪酬|薪金|补贴|薪|年终奖|年金|加班费|劳务费|劳务外包|提成|劳务派遣|绩效|酬劳|批量代付|PAYROLL|SALA|CPF|directors.*fees'
+     OR t2.CASH_TYPE REGEXP '代发|工资|劳务费'
+  );
+```
+
+---
+
+### CSV 第 10 行
+
+- 序号：`6.1`
+- 模型名称：`可疑财产`
+- 核心异常点（展示在前端页面）：`购房交易与房产登记不匹配`
+- 业务口径：`员工及关系人有购房交易，但名下房产无新增登记；有新增登记购房，但无相关购房交易记录。`
+- 相关指标：`/`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人及亲属`
+- 技术口径：`员工及其亲属账户中，存在购房支出，且在家庭负债表（ccdi_family_liability）中该员工名下无新增房产登记，则标记1`
+- 限制阈值指标：`/`
+- 可疑结果返回：`流水明细`
+- 风险等级：`空`
+- 原始代码：
+
+```sql
+--员工及其亲属购买房产但无资产登记
+select distinct bank_statement_id
+from
+(
+select t1.id_card
+,bank_statement_id
+,trx_time
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where project_id = PROJECT_ID
+and (user_memo rlike '(购|买).*房|房款|首付'
+or customer_account_name rlike '房地产|置业|置地|地产|房产|不动产|链家|贝壳|我爱我家|房管局')
+and amount_dr > 0
+union all
+select t1.person_id
+,bank_statement_id
+,trx_time
+from
+ccdi_staff_fmy_relation t1
+inner join
+ccdi_bank_statement t2
+on t1.relation_cert_no = t2.cret_no
+where t1.status = 1
+and t2.project_id = PROJECT_ID
+and (user_memo rlike '(购|买).*房|房款|首付'
+or customer_account_name rlike '房地产|置业|置地|地产|房产|不动产|链家|贝壳|我爱我家|房管局')
+and amount_dr > 0
+) t1
+left join
+(
+select person_id
+,max(updated_at) as updated_at
+from ccdi_asset_info
+where asset_main_type = 不动产    --根据具体数据确定码值
+and asset_sub_type in (住宅,商铺) --根据具体数据确定码值
+and asset_status = 正常 --根据具体数据确定码值
+group by person_id
+) t2
+on t1.id_card = t2.person_id
+where t1.trx_time > t2.updated_at --购买时间大于最近一次资产更新时间
+or t2.person_id is null; --未登记房产
+```
+
+**核对结论**
+
+- 原 SQL 引用了不存在的 `trx_time` 和 `updated_at`
+- 原 SQL 使用了未加引号的中文常量
+- 原技术口径提到 `ccdi_family_liability`，但当前库实际只有 `ccdi_asset_info`
+- 资产表中房产相关枚举值需按现网实际码值替换
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+SET @asset_main_type_house := 'REAL_ESTATE';
+SET @asset_sub_type_house_1 := 'HOUSE';
+SET @asset_sub_type_house_2 := 'SHOP';
+SET @asset_status_normal := 'NORMAL';
+
+SELECT DISTINCT
+    house_trade.bank_statement_id
+FROM (
+    SELECT
+        t1.id_card AS person_id,
+        t2.bank_statement_id,
+        STR_TO_DATE(LEFT(t2.TRX_DATE, 10), '%Y-%m-%d') AS trx_date
+    FROM ccdi_base_staff t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.id_card = t2.cret_no
+    WHERE t2.project_id = @project_id
+      AND (
+            t2.USER_MEMO REGEXP '(购|买).*房|房款|首付'
+         OR t2.CUSTOMER_ACCOUNT_NAME REGEXP '房地产|置业|置地|地产|房产|不动产|链家|贝壳|我爱我家|房管局'
+      )
+      AND t2.AMOUNT_DR > 0
+
+    UNION ALL
+
+    SELECT
+        t1.person_id,
+        t2.bank_statement_id,
+        STR_TO_DATE(LEFT(t2.TRX_DATE, 10), '%Y-%m-%d') AS trx_date
+    FROM ccdi_staff_fmy_relation t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.relation_cert_no = t2.cret_no
+    WHERE t1.status = 1
+      AND t2.project_id = @project_id
+      AND (
+            t2.USER_MEMO REGEXP '(购|买).*房|房款|首付'
+         OR t2.CUSTOMER_ACCOUNT_NAME REGEXP '房地产|置业|置地|地产|房产|不动产|链家|贝壳|我爱我家|房管局'
+      )
+      AND t2.AMOUNT_DR > 0
+) house_trade
+LEFT JOIN (
+    SELECT
+        person_id,
+        MAX(update_time) AS latest_asset_update_time
+    FROM ccdi_asset_info
+    WHERE asset_main_type = @asset_main_type_house
+      AND asset_sub_type IN (@asset_sub_type_house_1, @asset_sub_type_house_2)
+      AND asset_status = @asset_status_normal
+    GROUP BY person_id
+) asset
+    ON house_trade.person_id = asset.person_id
+WHERE asset.person_id IS NULL
+   OR house_trade.trx_date > DATE(asset.latest_asset_update_time);
+```
+
+**说明**
+
+- `REAL_ESTATE`、`HOUSE`、`SHOP`、`NORMAL` 只是占位值，必须替换成 `ccdi_asset_info` 的真实码值
+- 原业务口径还提到“有新增登记购房，但无相关购房交易记录”，该反向校验不在原 SQL 中，本次未擅自补充为双向模型
+
+---
+
+### CSV 第 11 行
+
+- 序号：`6.2`
+- 模型名称：`可疑财产`
+- 核心异常点（展示在前端页面）：`物业缴费与房产登记不匹配`
+- 业务口径：`员工及关系人有物业缴费记录，但名下房产无新增登记。`
+- 相关指标：`/`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人及亲属`
+- 技术口径：`员工及其亲属账户中，存在物业费支出，且在家庭负债表中无新增房产登记，则标记1`
+- 限制阈值指标：`/`
+- 可疑结果返回：`流水明细`
+- 风险等级：`空`
+- 原始代码：
+
+```sql
+有物业缴费记录但无房产登记
+select distinct bank_statement_id
+FROM
+(
+select t1.id_card
+,bank_statement_id
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where project_id = PROJECT_ID
+and (user_memo rlike '物业|物业费|管理费|物业服务|综合服务'
+or customer_account_name rlike '物业|小区|花园|苑|中心|大厦|业委会|业主委员会|置业|房地产|服务中心|管理处|社区')
+and amount_dr > 0
+union all
+select t1.person_id
+,bank_statement_id
+from
+ccdi_staff_fmy_relation t1
+inner join
+ccdi_bank_statement t2
+on t1.relation_cert_no = t2.cret_no
+where t1.status = 1
+and t2.project_id = PROJECT_ID
+and (user_memo rlike '物业|物业费|管理费|物业服务|综合服务'
+or customer_account_name rlike '物业|小区|花园|苑|中心|大厦|业委会|业主委员会|置业|房地产|服务中心|管理处|社区')
+and amount_dr > 0
+) t1
+left join
+(
+select person_id
+,max(updated_at) as updated_at
+from ccdi_asset_info
+where asset_main_type = 不动产
+and asset_sub_type in (住宅,商铺)
+and asset_status = 正常
+group by person_id
+) t2
+on t1.id_card = t2.person_id
+where t2.person_id is null;
+```
+
+**核对结论**
+
+- 原 SQL 引用了不存在的 `updated_at`
+- 原 SQL 的资产枚举值写法不符合当前库
+- 可完全改写为当前结构可执行版本
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+SET @asset_main_type_house := 'REAL_ESTATE';
+SET @asset_sub_type_house_1 := 'HOUSE';
+SET @asset_sub_type_house_2 := 'SHOP';
+SET @asset_status_normal := 'NORMAL';
+
+SELECT DISTINCT
+    property_fee.bank_statement_id
+FROM (
+    SELECT
+        t1.id_card AS person_id,
+        t2.bank_statement_id
+    FROM ccdi_base_staff t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.id_card = t2.cret_no
+    WHERE t2.project_id = @project_id
+      AND (
+            t2.USER_MEMO REGEXP '物业|物业费|管理费|物业服务|综合服务'
+         OR t2.CUSTOMER_ACCOUNT_NAME REGEXP '物业|小区|花园|苑|中心|大厦|业委会|业主委员会|置业|房地产|服务中心|管理处|社区'
+      )
+      AND t2.AMOUNT_DR > 0
+
+    UNION ALL
+
+    SELECT
+        t1.person_id,
+        t2.bank_statement_id
+    FROM ccdi_staff_fmy_relation t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.relation_cert_no = t2.cret_no
+    WHERE t1.status = 1
+      AND t2.project_id = @project_id
+      AND (
+            t2.USER_MEMO REGEXP '物业|物业费|管理费|物业服务|综合服务'
+         OR t2.CUSTOMER_ACCOUNT_NAME REGEXP '物业|小区|花园|苑|中心|大厦|业委会|业主委员会|置业|房地产|服务中心|管理处|社区'
+      )
+      AND t2.AMOUNT_DR > 0
+) property_fee
+LEFT JOIN (
+    SELECT
+        person_id
+    FROM ccdi_asset_info
+    WHERE asset_main_type = @asset_main_type_house
+      AND asset_sub_type IN (@asset_sub_type_house_1, @asset_sub_type_house_2)
+      AND asset_status = @asset_status_normal
+    GROUP BY person_id
+) asset
+    ON property_fee.person_id = asset.person_id
+WHERE asset.person_id IS NULL;
+```
+
+---
+
+### CSV 第 12 行
+
+- 序号：`6.3`
+- 模型名称：`可疑财产`
+- 核心异常点（展示在前端页面）：`大额纳税与资产登记不匹配`
+- 业务口径：`员工及关系人有5000元以上的纳税记录，但名下无房产车产新增登记。`
+- 相关指标：`/`
+- 指标英文名：`空`
+- 风险筛查对象：`员工本人及亲属`
+- 技术口径：`员工及其亲属账户中，存在单笔纳税 >5000 元，且在家庭负债表中无新增房产/车产登记，则标记1`
+- 限制阈值指标：`/`
+- 可疑结果返回：`流水明细`
+- 风险等级：`空`
+- 原始代码：
+
+```sql
+有5000元以上的纳税记录但无房产登记
+select distinct bank_statement_id
+FROM
+(
+select t1.id_card
+,bank_statement_id
+from
+ccdi_base_staff t1
+inner join
+ccdi_bank_statement t2
+on t1.id_card = t2.cret_no
+where project_id = PROJECT_ID
+and (user_memo rlike '税务|缴税|税款'
+or customer_account_name rlike '税务|税务局|国库|国家金库|财政')
+and amount_dr >= 5000
+union all
+select t1.person_id
+,bank_statement_id
+from
+ccdi_staff_fmy_relation t1
+inner join
+ccdi_bank_statement t2
+on t1.relation_cert_no = t2.cret_no
+where t1.status = 1
+and t2.project_id = PROJECT_ID
+and (user_memo rlike '税务|缴税|税款'
+or customer_account_name rlike '税务|税务局|国库|国家金库|财政')
+and amount_dr >= 5000
+) t1
+left join
+(
+select person_id
+,max(updated_at) as updated_at
+from ccdi_asset_info
+where asset_main_type = 不动产
+and asset_sub_type in (住宅,商铺)
+and asset_status = 正常
+group by person_id
+) t2
+on t1.id_card = t2.person_id
+where t2.person_id is null;
+```
+
+**核对结论**
+
+- 原 SQL 仍然只校验“无房产登记”，与业务口径中的“无房产车产新增登记”不一致
+- 原 SQL 引用了不存在的 `updated_at`
+- 资产类型应按真实码值配置
+
+**改写 SQL**
+
+```sql
+SET @project_id := 0;
+SET @tax_amount_floor := 5000;
+SET @asset_main_type_house := 'REAL_ESTATE';
+SET @asset_main_type_car := 'VEHICLE';
+SET @asset_status_normal := 'NORMAL';
+
+SELECT DISTINCT
+    tax_trade.bank_statement_id
+FROM (
+    SELECT
+        t1.id_card AS person_id,
+        t2.bank_statement_id
+    FROM ccdi_base_staff t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.id_card = t2.cret_no
+    WHERE t2.project_id = @project_id
+      AND (
+            t2.USER_MEMO REGEXP '税务|缴税|税款'
+         OR t2.CUSTOMER_ACCOUNT_NAME REGEXP '税务|税务局|国库|国家金库|财政'
+      )
+      AND t2.AMOUNT_DR >= @tax_amount_floor
+
+    UNION ALL
+
+    SELECT
+        t1.person_id,
+        t2.bank_statement_id
+    FROM ccdi_staff_fmy_relation t1
+    INNER JOIN ccdi_bank_statement t2
+        ON t1.relation_cert_no = t2.cret_no
+    WHERE t1.status = 1
+      AND t2.project_id = @project_id
+      AND (
+            t2.USER_MEMO REGEXP '税务|缴税|税款'
+         OR t2.CUSTOMER_ACCOUNT_NAME REGEXP '税务|税务局|国库|国家金库|财政'
+      )
+      AND t2.AMOUNT_DR >= @tax_amount_floor
+) tax_trade
+LEFT JOIN (
+    SELECT
+        person_id
+    FROM ccdi_asset_info
+    WHERE asset_status = @asset_status_normal
+      AND asset_main_type IN (@asset_main_type_house, @asset_main_type_car)
+    GROUP BY person_id
+) asset
+    ON tax_trade.person_id = asset.person_id
+WHERE asset.person_id IS NULL;
+```
+
+**说明**
+
+- 该改写版按业务口径补充了“车产”主类判断，但主类码值仍需按现网真实枚举替换
+
+---
+
+## 汇总结论
+
+- CSV 共 `12` 行，其中：
+  - `10` 行带 `sql<br>` 前缀 SQL
+  - `1` 行带原始 SQL 但未写 `sql<br>` 前缀
+  - `1` 行仅用于补充阈值参数，没有独立 SQL
+- 当前库能直接承接并改写的模型：
+  - `2.2`
+  - `3.1`
+  - `3.2`
+  - `4`
+  - `5.1`
+  - `5.2`
+  - `CSV 第 9 行补充规则`
+  - `6.1`
+  - `6.2`
+  - `6.3`
+- 当前库不能完全独立落地的模型：
+  - `2.1`
+
+## 语法校验结果
+
+- 已使用 `EXPLAIN` 在当前 `MySQL 5.7.44 / ccdi` 上校验通过的改写 SQL：
+  - `2.2`
+  - `3.1`
+  - `3.2`
+  - `4`
+  - `5.1`
+  - `5.2`
+  - `CSV 第 9 行补充规则`
+  - `6.1`
+  - `6.2`
+  - `6.3`
+- 未纳入自动语法校验的改写 SQL：
+  - `2.1`
+  - 原因：改写版依赖 `tmp_model_21_loan_cust_acct`、`tmp_model_21_assure_cust_acct`、`tmp_model_21_intermediary_acct` 三张引用表，当前库尚未创建
+
+## 建议后续动作
+
+1. 为 `2.1` 建立客户账号引用清单表，或者把外部客户库同步到 `ccdi`
+2. 为资产模型补齐 `asset_main_type`、`asset_sub_type`、`asset_status` 的字典码值说明
+3. 将 `PROJECT_ID`、上下限金额等参数沉淀到模型参数表，而不是保留在文本 SQL 中
+4. 如果后续要在代码里执行这些 SQL，建议统一封装成 MyBatis XML，并配套 DTO/VO 与单元测试