Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股机器学习因子挖掘/自动化因子发现。当用户说"因子挖掘"、"ML因子"、"机器学习因子"、"自动因子发现"、"factor mining"、"遗传规划因子"时触发。基于 cn-stock-data 获取数据，使用ML方法自动挖掘有效因子。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-ml-factor-mining/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -26% | 0% |
通过 cn-stock-data skill 获取数据：
# 机器学习因子挖掘报告
## 一、因子空间
| 候选因子数 | 有效因子数 | 筛选率 |
|-----------|-----------|--------|
## 二、Top因子表现
[因子表达式、IC、IR、多空收益]
## 三、因子组合
[最终因子池、权重分配]
## 四、过拟合检验
[样本外表现、稳定性评估]## 因子挖掘速览
- 候选因子 5,000 个，筛选出 18 个有效因子
- Top因子 IC=0.05, IR=1.2
- 样本外多空年化 15.3%
- 过拟合风险：低(多重检验通过)参考 references/ml-factor-mining-guide.md 获取详细方法论与 A股实证研究。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 31,612 | 29,934 | -5% | 1 | 1 | 0% | 5,179 | 5,667 | +9% | 0 | 0 | — |
case-02 | fail→pass | 20,198 | 19,787 | -2% | 1 | 1 | 0% | 3,198 | 2,840 | -11% | 0 | 0 | — |
case-03 | fail→fail | 43,370 | 38,658 | -11% | 1 | 1 | 0% | 6,214 | 6,199 | -0% | 0 | 0 | — |
case-04 | fail→fail | 29,333 | 20,134 | -31% | 1 | 1 | 0% | 3,790 | 3,741 | -1% | 0 | 0 | — |
case-05 | pass→pass | 17,717 | 17,715 | -0% | 1 | 1 | 0% | 2,597 | 2,914 | +12% | 0 | 0 | — |
case-06 | pass→pass | 19,506 | 25,475 | +31% | 1 | 1 | 0% | 2,797 | 3,850 | +38% | 0 | 0 | — |
case-07 | fail→fail | 18,845 | 16,499 | -12% | 1 | 1 | 0% | 2,724 | 2,907 | +7% | 0 | 0 | — |
case-08 | fail→pass | 18,979 | 20,018 | +5% | 1 | 1 | 0% | 2,833 | 3,316 | +17% | 0 | 0 | — |
case-14 | pass→pass | 20,482 | 19,620 | -4% | 1 | 1 | 0% | 3,120 | 3,482 | +12% | 0 | 0 | — |
case-09 | fail→pass | 21,069 | 23,852 | +13% | 1 | 1 | 0% | 3,475 | 3,887 | +12% | 0 | 0 | — |
case-10 | pass→pass | 25,707 | 22,357 | -13% | 1 | 1 | 0% | 3,528 | 3,663 | +4% | 0 | 0 | — |
case-11 | pass→pass | 22,285 | 15,061 | -32% | 1 | 1 | 0% | 2,757 | 3,025 | +10% | 0 | 0 | — |
case-12 | pass→pass | 52,947 | 21,689 | -59% | 1 | 1 | 0% | 3,189 | 3,264 | +2% | 0 | 0 | — |
case-13 | pass→pass | 22,311 | 20,422 | -8% | 1 | 1 | 0% | 2,852 | 4,008 | +41% | 0 | 0 | — |
case-15 | fail→pass | 18,005 | 5,304 | -71% | 1 | 1 | 0% | 2,501 | 1,333 | -47% | 0 | 0 | — |
case-16 | fail→pass | 17,737 | 10,523 | -41% | 1 | 1 | 0% | 2,393 | 1,767 | -26% | 0 | 0 | — |
case-17 | pass→pass | 18,580 | 20,659 | +11% | 1 | 1 | 0% | 3,141 | 3,367 | +7% | 0 | 0 | — |
case-18 | fail→pass | 10,710 | 5,424 | -49% | 1 | 1 | 0% | 1,594 | 1,386 | -13% | 0 | 0 | — |
case-19 | pass→pass | 12,284 | 6,147 | -50% | 1 | 1 | 0% | 1,495 | 1,441 | -4% | 0 | 0 | — |
case-20 | pass→pass | 22,361 | 21,075 | -6% | 1 | 1 | 0% | 3,022 | 3,585 | +19% | 0 | 0 | — |
case-21 | fail→fail | 47,516 | 34,071 | -28% | 1 | 1 | 0% | 4,306 | 4,949 | +15% | 0 | 0 | — |
case-22 | fail→fail | 31,387 | 32,260 | +3% | 1 | 1 | 0% | 4,944 | 6,349 | +28% | 0 | 0 | — |
case-23 | fail→fail | 23,466 | 30,037 | +28% | 1 | 1 | 0% | 3,819 | 4,865 | +27% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.