financial-analyst-data-lite

EN: A-share historical OHLCV + valuation + financials + TDX F10 events, packaged in Qlib binary + Parquet formats. Companion dataset for financial-analyst — a 14-agent single-stock deep-dive research workstation.

中文: A 股历史行情 + 估值 + 财报 + TDX F10 事件数据集, Qlib 二进制 + Parquet 双格式打包. 配套 financial-analyst — 14 Agent 个股深度研究工作站使用.

Published / 发布: 2026-05-24 · Size / 体量: ~2.72 GB · License: Apache 2.0

---

📊 Three Preset Tiers / 三档预设

Pick the tier that fits your use case. / 按需选择合适档位.

	demo	lite	full
Size / 体量	~155 MB	~3 GB	~14 GB
Stocks / 股票池	300 (current CSI300 by mv)	800 (CSI800 ≈)	5500+ (all A-share incl. delisted)
Daily OHLCV+估值	✅	✅	✅
5min OHLCV	❌	✅ (~7 days)	✅
Financial reports / 财务报表	❌	✅ (735 MB)	✅
F10 text / F10 原始文本	❌	✅ (1323 codes)	✅
TDX 历年财报 zip	❌	❌	✅ (257 MB)
Best for / 适合	试用 / try-out	fa report 研报 / multi-stock research	量化研究 / quant research
HF Repo	data-demo	data-lite	data-full

This repo is the lite tier. / 此 repo 是 lite 档.

---

📦 What's Included / 数据清单

English

• 800 stocks daily OHLCV + 7 valuation fields (PE / PB / PS / DV / MV / CIRC_MV / turnover_rate)
• Date range (daily): 1990-12-19 → 2026-12-31 (8797 trading days)
• 5min OHLCV ✅ (800 stocks, 1.30 GB)
• Financial reports (parquet/financial/) ✅ (24 parquet files total, 750.0 MB including all parquet)
• F10 text (news_data/tdx_f10/) ✅ (7363 .txt files, 294.0 MB)
• TDX historical financial zip (tdx_finance/) ❌ (0 zip files, 0.0 MB)

中文

• 800 只股票 日线 OHLCV + 7 个估值字段 (市盈率/市净率/市销率/股息率/总市值/流通市值/换手率)
• 日线日期范围: 1990-12-19 → 2026-12-31 (8797 个交易日)
• 5min 行情 ✅ (800 只, 1.30 GB)
• 完整财务报表 (parquet/financial/) ✅ (24 个 parquet, 共 750.0 MB)
• F10 原始文本 (news_data/tdx_f10/) ✅ (7363 个 .txt, 294.0 MB) — 公司大事 / 龙虎榜单 / 主力追踪 / 最新提示
• TDX 历年财报 zip (tdx_finance/) ❌ (0 个 zip, 0.0 MB) — 用 scripts/import_tdx_financial.py 解

---

🗂 Directory Layout / 目录布局

cn_data/                          # daily — Qlib binary
  calendars/day.txt               # trading calendar / 交易日历
  instruments/all.txt             # 800 codes
  features/{code}/              # one dir per stock, e.g. sh600519/
    open.day.bin                  # [4-byte float32 start_idx] + [float32 array]
    high.day.bin
    low.day.bin
    close.day.bin                 # 不复权收盘 / unadjusted close
    volume.day.bin                # 手 / lots (= 100 shares)
    amount.day.bin                # 元 / CNY
    pe_ttm.day.bin                # TTM PE ratio
    pb.day.bin
    ps_ttm.day.bin                # may be NaN for some history
    dv_ttm.day.bin                # dividend yield %, may be NaN
    total_mv.day.bin              # 万元 / 10K CNY
    circ_mv.day.bin               # 流通市值万元 / circulating mv (10K CNY)
    turnover_rate.day.bin         # %
cn_data_5min/                     # 5-minute — same Qlib binary layout, rolling ~7 days
parquet/                          # 非时序结构化数据 / non-time-series structured data
  industry_boards.parquet         # 同花顺一级行业 / 10jqka level-1 industry
  index_constituents.parquet      # CSI300 / CSI500 成份 / constituents
  tdx_f10_index.parquet           # F10 事件索引 (公司大事/龙虎榜/研报)
  tdx_f10_warnings_latest.parquet # 最新负向预警 / latest warning events
  northbound_holding.parquet      # 北向资金持仓 / northbound stake
  tushare_stock_basic.parquet     # 股票基本信息 / basic listing info
  concept_ths_*.parquet           # 同花顺概念 / 10jqka concept boards
  events/                         # 公司公告 / company filings
  institutional/                  # 机构持仓 / institutional holders
  blocks/                         # 板块映射 / sector mappings
  xdxr/                           # 分红除权 / dividends + splits
  financial/                      # 完整财务报表 (~735MB) — 资产/负债/利润/现金流/指标

news_data/tdx_f10/{{code}}/         # F10 原始 .txt — 跟 tdx_f10_index 配合用

---

🚀 Usage / 使用方法

Option 1 — via financial-analyst CLI (recommended / 推荐)

EN: Easiest. The CLI handles download, path setup, and integration.

中文: 最简. CLI 自动下载、配置路径、衔接 agent.

pip install financial-analyst

# Interactive wizard, picks this dataset / 交互向导自动用本数据集
fa init

# Generate a deep-dive research report on 茅台 / 跑研报
fa report SH600519

Option 2 — Direct download + Qlib

EN: Pull the dataset with huggingface_hub, then use Qlib's D.features() API.

中文: 用 huggingface_hub 下载, 用 Qlib D.features() API 读数据.

from huggingface_hub import snapshot_download

local_dir = snapshot_download(
    repo_id="yifishbossman/financial-analyst-data-lite",
    repo_type="dataset",
    local_dir="~/.financial-analyst/data",
)

import qlib
from qlib.data import D
qlib.init(provider_uri=f"{local_dir}/cn_data", region="cn")

df = D.features(
    ["SH600519"], ["$close", "$volume", "$pe_ttm", "$total_mv"],
    start_time="2024-01-01", end_time="2026-05-31", freq="day",
)
print(df.tail())

Option 3 — Read Parquet directly with pandas

EN: Non-time-series data (financials / industry / F10 / events) is plain Parquet.

中文: 非时序数据 (财报 / 行业 / F10 / 事件) 是普通 Parquet, pandas 直接读.

import pandas as pd

# Industry classification for all listed stocks / 全市场行业分类
ind = pd.read_parquet(f"{local_dir}/parquet/tushare_stock_basic.parquet")

# Latest TDX F10 negative warnings (last 7 days) / TDX F10 最新 7 天负向事件
warn = pd.read_parquet(f"{local_dir}/parquet/tdx_f10_warnings_latest.parquet")

# CSI300 / 500 constituents / 沪深 300/500 成份
idx = pd.read_parquet(f"{local_dir}/parquet/index_constituents.parquet")

---

📐 Units & Conventions / 单位与约定

Field / 字段	Unit / 单位	Notes / 说明
open / high / low / close	元 / CNY	not adjusted / 不复权 — adjustment factor not included
volume	手 (= 100 股) / lots	Tushare convention; NOT pytdx convention (股 / shares)
amount	元 / CNY	成交额 / turnover value
pe_ttm / pb / ps_ttm	(无单位) / dimensionless	TTM ratios / 滚动 12 月
dv_ttm	%	dividend yield TTM / 股息率
total_mv / circ_mv	万元 / 10K CNY	total / circulating market cap / 总/流通市值
turnover_rate	%	daily turnover ratio / 换手率

---

🔗 Data Sources & Lineage / 数据来源

English

• OHLCV + valuation: Tushare Pro (pro.daily + daily_basic), HTTP endpoint, ~5500 A-share tickers
• 5min OHLCV: TDX main sites via pytdx (free, no token required for historical 5min)
• Financial reports: Tushare Pro (fina_indicator, income, balancesheet, cashflow)
• TDX F10 events: pytdx direct connection to broker hosts (招商证券/东兴/华泰 etc.), parsed company events / 龙虎榜 / institutional flows
• For daily updates (post-download): users run fa data update which pulls from pytdx main sites (free, no token) + Tencent qt.gtimg.cn realtime (free, no cookie). See direct-data-stability research.

中文

• 日线 OHLCV + 估值: Tushare Pro (pro.daily + daily_basic) HTTP 接口, 覆盖全 A 股约 5500 只
• 5min 行情: TDX 主站经 pytdx 拉取 (零成本, 历史 5min 不需 token)
• 财务报表: Tushare Pro (fina_indicator, income, balancesheet, cashflow)
• TDX F10 事件: pytdx 直连券商主站 (招商证券/东兴/华泰 等), 解析公司大事 / 龙虎榜 / 主力追踪
• 下载后日常更新: 用户跑 fa data update, 走 pytdx 主站 (免 token) + 腾讯 qt.gtimg.cn 实时 (免 cookie). 详见 直连数据稳定性研究.

---

⚠️ Disclaimer / 免责声明

EN: This dataset is provided strictly for research and educational purposes. Data accuracy is not guaranteed; verify independently before any trading decision. The publisher assumes no liability for losses incurred from use of this data. Redistribution must comply with original source terms — see Tushare ToS.

中文: 本数据集仅供学术研究 / 教学使用, 不保证数据准确性. 任何投资决策须自行独立验证, 因使用本数据造成的损失发布方概不负责. 二次分发须遵守原始数据源条款 — 参考 Tushare 服务协议.

---

📄 License / 许可

• Code / 工具脚本: Apache 2.0 (financial-analyst toolchain)
• Data / 数据本体: 遵从原始数据源 (Tushare / TDX) 各自的使用条款 / refer to original sources
• Citation / 引用: If you use this dataset in a paper, please cite financial-analyst / 论文引用请标注 financial-analyst

---

🔄 Updating This Dataset / 数据更新

EN: This snapshot is static. For incremental daily updates (post-market each day), install the financial-analyst package locally:

中文: HF 上的快照是静态的. 每天盘后增量更新, 本地装 financial-analyst 包:

pip install financial-analyst
fa data update           # incremental day OHLCV + valuation via pytdx (free)
fa data update --5min    # incremental 5min via TDX local client
fa data update --f10     # refresh TDX F10 events for watched stocks

---

🤝 Contributing / 贡献

EN: Issues / PRs welcome on the main repo. For dataset-specific issues (missing codes, schema questions), file an issue tagged dataset.

中文: Bug 反馈 / 功能建议请去 主仓库. 数据集相关问题 (代码缺失 / schema 问题) 请加 dataset 标签.

---

_{Generated by scripts/publish_hf_dataset.py on 2026-05-24 · v1.9.6 · bilingual zh/en}