emd
Explain My Data — instant EDA reports from any CSV or Excel file.
Why
Every new dataset starts with the same half-hour of boilerplate: check distributions, scan for missing values, eyeball correlations, hunt for outliers. emd runs that entire first pass in one command and turns it into a single Markdown report with charts, so you can skip straight to the interesting questions. It's built for data scientists and analysts who want a fast, repeatable first look at a file before opening a notebook. It also detects statistical drift between two datasets, which makes it useful for monitoring train/test splits or production data in an MLOps pipeline. No config, no setup beyond installing it — point it at a file and read the report.
Install
git clone https://github.com/utkukosman1/explain-my-data-CLI.git cd explain-my-data-CLI pip install -e .
For Isolation Forest outlier detection (optional):
Quickstart
Produces reports/<filename>/report.md with embedded charts — distributions, correlations, missing values, outliers, and more, all in one file.
What you get
| Section | What you get |
|---|---|
| Data Quality Summary | FATAL/WARNING/INFO checks — empty data, high missing columns, duplicate rows, mixed types |
| Dataset Overview | Shape, dtypes, null counts, memory usage |
| Distribution Analysis | Mean, median, std, CV, skewness, kurtosis, normality test, percentiles, histogram + KDE, box plot |
| Correlation Analysis | Pearson & Spearman heatmaps, Cramér's V (categorical), point-biserial, VIF, strongest pairs table |
| Missing Value Analysis | Per-column missing %, missingness patterns, correlated missingness pairs |
| Outlier Detection | IQR, Z-score, Modified Z-score, optional Isolation Forest |
| Key Insights (Target) | Top correlated features + hue charts — only when --target is provided |
More commands
emd summary data.csv # fast terminal recap — no files written emd doctor data.csv # health score + pre-modeling risk audit emd analyze titanic.csv --target Survived # feature importance vs. a target column emd compare train.csv test.csv # drift detection between two datasets emd check data.csv # quality check only, no report emd batch ./data/ # analyze every CSV/XLSX in a directory emd schema init data.csv # generate a YAML schema contract
See docs/cli-reference.md for the full flag reference, and docs/statistical-methods.md for how each statistic and test is computed.
Optional: web UI
A small companion UI (FastAPI + Next.js) exists for browser use — see web/. It's not required to use emd.
Development
pip install -e ".[dev]"
pytest tests/
ruff check src/ tests/
mypy src/emdLicense
MIT — see LICENSE.