fenrix-pm-decision-news-staging
收藏资源简介:
FENRIX PM Decision News是一个专为学术研究人员(教授)设计的金融新闻数据集,主要用于投资组合经理决策研究。数据集根据文章意义进行组织,支持文本分类和特征提取等自然语言处理任务。当前版本(teammate切片)包含来自2026年7月teammate事件包的已验证数据,覆盖2019年7月1日至2026年6月30日期间的数据,总计44,904行,存储在656个Parquet文件中(其中572个为新闻文件,84个为结构化数据文件)。数据源多样,包括Guardian、IMF、FRED、Federal Register、GDELT、Census、EIA、BEA、FDA、CFPB、FDIC和BLS等权威机构。数据集主要分为两部分:新闻数据(按主题分类为宏观经济、行业板块、单只股票、多公司背景和市场背景等文件夹)和结构化数据(官方经济/监管序列)。包含41个字段,关键字段有文章ID、文章组、文章类型、主题、市场范围、发布时间、事件日期、标题、描述、正文、股票代码、公司、行业板块、宏观标签、发布者、许可类别和再分发状态等。需要注意的是,该版本不包含原始的ZIP、JSONL或TXT摄入文件,也不包含完整的仓库数据(37.57M行)。数据许可因来源而异,使用前需检查行级许可信息,并附带了详细的QA报告以供验证。
FENRIX PM Decision News is a financial news dataset designed for academic researchers (professors), specifically for portfolio manager decision research. Its content is organized based on article meaning to support natural language processing tasks such as text classification and feature extraction. The current released version (teammate slice) contains validated data from the July 2026 teammate event package, covering the period from July 1, 2019, to June 30, 2026. The data scale is 44,904 rows, stored in 656 Parquet files (572 news files and 84 structured data files). Data sources are diverse, including authoritative institutions such as Guardian, IMF, FRED, Federal Register, GDELT, Census, EIA, BEA, FDA, CFPB, FDIC, and BLS. The dataset mainly consists of two parts: news data (categorized into folders by topic such as macroeconomics, industry sectors, single stocks, multi-company context, and market context) and structured data (official economic/regulatory sequences). It includes 41 fields, with key fields such as article ID, article group, article type, topic, market scope, publication time, event date, title, description, body, ticker, company, industry sector, macro tags, publisher, license category, and redistribution status. Note that this version does not include original ZIP, JSONL, or TXT ingestion files, nor the full repository data (37.57M rows). Data licenses vary by source, and row-level license information should be checked before use. The dataset comes with a detailed QA report for validation.
数据集概述:FENRIX PM Decision News
简要描述:
面向投资组合经理决策研究的金融新闻产品,按文章语义组织,当前发布版本为已验证的团队成员产品切片(teammate product slice)。
许可证:MIT
任务类别:文本分类、特征提取
语言:英语
数据集大小:约44,904行(10K < n < 100K)
配置:
teammate_news:新闻数据,Parquet格式,路径Data/News/**/*.parquetteammate_structured:结构化数据,Parquet格式,路径Data/Structured/**/*.parquet
当前发布范围
| 指标 | 数值 |
|---|---|
| 已接受行数 | 44,904 |
| 日期覆盖范围 | 2019-07-01 至 2026-06-30 |
| Parquet文件总数 | 656 |
| Data/News Parquet文件 | 572 |
| Data/Structured Parquet文件 | 84 |
数据来源
- Guardian, IMF, FRED, Federal Register, GDELT, Census, EIA, BEA, FDA, CFPB, FDIC, BLS 等新闻行。
未包含内容
- 原始zip、jsonl或txt intake文件
- 完整仓库支持的产品构建(37,571,131行银层数据)——计划作为独立暂存构建
- API密钥或导出的Parquet中的本地文件系统路径
文件夹布局
| 文件夹 | 内容 |
|---|---|
Data/News/Economic_Macro/ |
宏观发布、利率、GDP、全球宏观 |
Data/News/Sector_Industry/ |
能源、行业主题 |
Data/News/Single_Stock/ |
公司特定Guardian和标记行 |
Data/News/Multi_Company/ |
多公司上下文 |
Data/News/Market_Context/ |
监管、银行、地缘政治、未分类上下文 |
Data/Structured/ |
官方结构化经济/监管序列 |
Reports/ |
团队成员摄取对账与QA报告 |
Docs/ |
产品说明与教授指南 |
验证说明
使用Hugging Face API文件树和 Reports/Teammate_* 文件作为此发布的事实依据。不依赖数据集查看器预览,因其可能显示早期暂存状态的过时样本。
必需QA报告:
Reports/Teammate_Ingest_Summary.csvReports/Teammate_Source_Row_Counts.csvReports/Teammate_Date_Coverage.csvReports/Teammate_Product_Path_Counts.csvReports/Teammate_Article_Group_Counts.csvReports/Teammate_Full_Text_Join_Report.csvReports/Teammate_Duplicate_Report.csvReports/Teammate_Secret_Scan_Report.csvReports/Teammate_Guardian_Ingest_Summary.csvReports/Teammate_Staging_Upload_Manifest.csvReports/Professor_Summary_Teammate_Update.mdDocs/Teammate_Data_Notes.md
快速加载示例
python import pandas as pd from huggingface_hub import hf_hub_download
path = hf_hub_download( repo_id="Scottswi/fenrix-pm-decision-news", filename="Data/News/Economic_Macro/Global_Macro/Imf/year=2020/month=01/part-imf-2020-01-00000.parquet", repo_type="dataset", ) df = pd.read_parquet(path) print(df.columns.tolist()) print(len(df))
模式(41列)
关键字段:article_id, article_group, article_type, topic, market_scope, published_at, event_date, title, description, body, body_available, tickers, companies, sectors, macro_tags, canonical_url, publisher, source_id, license_class, redistribution_status, provenance_uri, source_record_id, source_package, ingested_at_utc。
许可与转载
来源有多种条款。在外部共享前,请使用行级 license_class 和 redistribution_status。Guardian及部分新闻行可能为 restricted_or_unknown 或需非商业审查。官方宏观/监管结构化行通常为公共领域,需注明出处。
相关数据集




