# Gru

> 使用 GRU 对足量时间序列进行预测（forecast deep learning）

- Skill: `zafer-liu/gru` (Agent Skill)
- Install (CLI): `npx skillmds@latest add zafer-liu/gru`
- Raw SKILL.md: https://api.skillmd.com/api/skills/zafer-liu/gru/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: Zafer-Liu (https://skillmd.com/u/zafer-liu)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/zafer-liu/gru

---

# GRU 预测

仅在样本量和序列长度足够时使用。确认窗口、特征和预测期，严格按时间划分训练验证，避免泄漏，报告基线对比、误差和不确定性；小数据优先建议传统时序模型。

## Tool routing

1. Use `get_schema` to identify the time column, target column, feature columns, and source table.
2. Use `query_data` only to verify sorted sequence length, missingness, and whether the sample size is sufficient.
3. Use `run_analysis` with `analysis_name="Time_Series_GRU"` for the actual model computation.
4. Use `generate_chart` on forecast or validation result tables after `run_analysis` succeeds.

## Implementation reference

- Tool entry: `agent/tools/business/data.py::_tool_run_analysis`
- Analysis registry: `Function/Analyze/registry.py`
- Analysis implementation: `Function/Analyze/Time_Series_GRU/analyze.py`
- Chart implementation: `Function/Charts_generation/chart_generate.py`