maic-results
收藏资源简介:
MAIC 数据集是早期预警系统管道的配套数据仓库,包含结果表格、HMM 标签和推理日志。该数据集汇集了来自管道各个阶段的输出,包括基线模型、消融实验和生产模型的结果(07_aggregation)、基于 HMM 发作的提前时间分析(09_lead_time)、HMM 标签一致性及 XGBoost 对比检查(10_hmm_robustness)、全局与局部 HMM 稳定性及三阶段危机验证(11_external_validation)、持久性/多数基线及外部文档化危机时间戳的提前时间分析(13_baselines_bootstrap)、实时推理与结果日志(12_inference_logs)以及全局 HMM 压力标签(labels/)。标签数据以每资产每月10秒分辨率提供。该数据集旨在为 MAIC 管道的结果提供标准化、可复用的公共副本,适用于时间序列预测、危机预警、模型评估与验证等研究场景。
The MAIC dataset is a companion data repository for the early warning system pipeline, containing result tables, HMM labels, and inference logs. It aggregates outputs from various stages of the pipeline, including baseline models, ablation experiments, and production model results (07_aggregation), HMM-based lead time analysis (09_lead_time), HMM label consistency and XGBoost comparison checks (10_hmm_robustness), global and local HMM stability with three-phase crisis validation (11_external_validation), persistence/majority baseline and externally documented crisis timestamp lead time analysis (13_baselines_bootstrap), real-time inference and result logs (12_inference_logs), and global HMM stress labels (labels/). The label data is provided at a resolution of 10 seconds per asset per month. This dataset aims to provide a standardized, reusable public copy of the MAIC pipeline results, suitable for research scenarios such as time series forecasting, crisis early warning, model evaluation, and validation.
数据集概述
该数据集是MAIC早期预警系统管道的配套数据仓库,包含结果表格、HMM标签和推理日志。
数据集内容
数据集主要包含以下目录和内容:
07_aggregation/:基线模型、消融模型和生产模型的结果(07a-07c)09_lead_time/:针对HMM发作的提前期分析(09)10_hmm_robustness/:HMM标签一致性及XGBoost对比检查(10)11_external_validation/:全局与局部HMM稳定性、三级危机验证、与外部验证者的Cohens kappa一致性(11a-11b)13_baselines_bootstrap/:持续性/多数类基线、针对外部记录的危机时间戳的提前期分析、块自举稳健性检查(13a-13c)12_inference_logs/:实时推理和结果日志(如果存在)labels/:全局HMM压力标签,按资产、按月、10秒分辨率
背景说明
这些表格此前仅存在于运行管道的机器上,本仓库是其规范的公开副本。
相关链接
- 代码:https://github.com/Goodie-Goody/maic
- 模型:https://huggingface.co/Goooddy/maic-models
- DOI:https://doi.org/10.57967/hf/9683
许可证
该数据集采用MIT许可证。



