Datasetpublic

ismailtasdelen/global-asset-market-cap-intelligence

Global Asset Market Capitalization Intelligence Dataset (GAMCID) What is GAMCID? GAMCID is a research-grade, machine-learning-ready dataset capturing the historical evolution of global assets ranked by market capitalization. It covers public companies, precious metals, cryptocurrencies, ETFs, and commodities — sourced from CompaniesMarketCap.com. Why does it exist? Existing financial datasets typically focus on single asset classes (stocks OR crypto… See the full description on the dataset page: https://huggingface.co/datasets/ismailtasdelen/global-asset-market-cap-intelligence.

sourceHugging Faceotherupdated 17d agoView on Hugging Face

0likes72downloads

Dataset Card

Global Asset Market Capitalization Intelligence Dataset (GAMCID)

What is GAMCID?

GAMCID is a research-grade, machine-learning-ready dataset capturing the historical evolution of global assets ranked by market capitalization. It covers public companies, precious metals, cryptocurrencies, ETFs, and commodities — sourced from CompaniesMarketCap.com.

Why does it exist?

Existing financial datasets typically focus on single asset classes (stocks OR crypto OR commodities). GAMCID unifies them into a single ranked universe, enabling:

—Cross-asset-class analysis (Gold vs Tech Companies, Bitcoin vs S&P 500 ETFs)
—Global ranking dynamics and transition events
—Machine learning tasks: ranking prediction, top-N entry/exit, momentum
—Sector, country, and asset-type dominance tracking over time

How is data collected?

1.Listing page: The top-N assets are scraped from /assets-by-market-cap/ (paginated, 100 per page)
2.Individual asset pages: Each asset's detail page embeds full historical market cap data (daily, embedded as data = [{d: timestamp, m: value}] in the HTML)
3.Conservative rate limiting: 2-second delays between requests, exponential backoff retries, respects robots.txt

Data is never fabricated. Missing values remain as null. Historical observations are real scraped data points.

Installation

bash

git clone https://github.com/ismailtasdelen/global-asset-market-cap-dataset
cd global-asset-market-cap-dataset
pip install -r requirements.txt

Reproduction

bash

# Full pipeline
python scripts/collect.py          # Phase 1: Collect raw data
python scripts/process.py          # Phase 2: Normalize + feature engineering
python scripts/validate.py         # Phase 3: Validate + quality report
python scripts/build_dataset.py    # Phase 4: Export Parquet/CSV/JSONL

# Or use make:
make all

Dataset Schema

Column	Type	Description
`date`	date	Observation date (UTC)
`asset_id`	string	Stable identifier (e.g., `NVDA_US`, `GOLD_METAL`)
`asset_name`	string	Normalized asset name
`asset_type`	string	Company, Cryptocurrency, Precious Metal, ETF, Commodity, Other
`rank`	int	Global market cap rank on this date
`market_cap_usd`	float	Market capitalization in USD
`price_usd`	float	Price per unit in USD (when available)
`country`	string	Country of origin/headquarters
`sector`	string	Sector classification
`daily_change_pct`	float	Daily percentage change
`rank_change_1d/7d/30d/90d/1y`	int	Rank improvement vs prior periods
`market_cap_growth_*_pct`	float	Market cap growth percentage
`momentum_*`	float	Market cap momentum indicator
`global_power_score`	float	Derived composite score (0-1)
`is_top_10/25/50/100/250/500/1000`	boolean	Top-N membership flags
`future_rank_`, `future_market_cap_`		Leakage-safe prediction targets

Feature Definitions

—Rank Change: previous_rank - current_rank. Positive = rank improved (numerically lower = better).
—Momentum: (market_cap_current / market_cap_previous_period) - 1
—Global Power Score: Weighted composite of normalized market cap (0.50), rank (0.20), growth (0.15), momentum (0.15). Configurable in config/config.yaml.
—Future Targets: Calculated strictly from future observations. No data leakage.

Known Limitations

—Historical depth varies per asset (some go back to 1999, newer assets have shorter history)
—Precious metal market caps are estimates per the source
—Price data is not available for all assets on all dates
—Country/sector coverage is incomplete for some asset types
—The dataset reflects only publicly visible market cap data from the source

License

See LICENSE. Data sourced from CompaniesMarketCap.com. This dataset is for research and educational purposes only — not investment advice.

Citation

Ismail Tasdelen.
Global Asset Market Capitalization Intelligence Dataset.
Hugging Face Datasets.
2026.

Architecture

mermaid

flowchart LR
    A[CompaniesMarketCap.com] --> B[Collector]
    B --> C[Raw Data]
    C --> D[Normalization]
    D --> E[Feature Engineering]
    E --> F[Validation]
    F --> G[Parquet / CSV / JSONL]
    G --> H[Hugging Face Dataset]