gru
GitHub基于GRU深度学习模型进行时间序列预测。通过校验数据充足性,严格按时间划分训练验证集,执行模型计算并生成可视化图表,同时提供基线对比与不确定性评估。
触发场景
需要进行时间序列预测
使用深度学习模型分析时序数据
安装
npx skills add Zafer-Liu/Data-Analysis-Agent --skill gru -g -y
SKILL.md
Frontmatter
{
"icon": "🧠",
"name": "gru",
"description": "使用 GRU 对足量时间序列进行预测(forecast deep learning)",
"allowedTools": [
"get_schema",
"query_data",
"run_analysis",
"generate_chart"
]
}
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
- Use
get_schemato identify the time column, target column, feature columns, and source table. - Use
query_dataonly to verify sorted sequence length, missingness, and whether the sample size is sufficient. - Use
run_analysiswithanalysis_name="Time_Series_GRU"for the actual model computation. - Use
generate_charton forecast or validation result tables afterrun_analysissucceeds.
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
版本历史
- d6a2c3e 当前 2026-07-24 12:12


