Metadata-Version: 2.4
Name: athena-eda
Version: 0.2.2
Summary: Automated EDA, ML readiness scoring, and data quality checks — locally in your browser.
Author: Athena
License: MIT
Project-URL: Homepage, https://athena-alpha-hazel.vercel.app
Project-URL: Repository, https://github.com/SaurabhLingam/Athena
Keywords: eda,machine-learning,data-science,exploratory-data-analysis
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: fastapi>=0.111.0
Requires-Dist: uvicorn[standard]>=0.29.0
Requires-Dist: pandas>=2.0.0
Requires-Dist: numpy>=1.24.0
Requires-Dist: scikit-learn>=1.3.0
Requires-Dist: lightgbm>=4.3.0
Requires-Dist: nltk>=3.8.0
Requires-Dist: python-dotenv>=1.0.0
Requires-Dist: slowapi>=0.1.9
Requires-Dist: python-multipart>=0.0.9
Requires-Dist: scipy>=1.11.0
Requires-Dist: rich>=13.7.0
Requires-Dist: typer>=0.12.0
Requires-Dist: starlette>=0.27.0
Requires-Dist: statsmodels>=0.14.0
Requires-Dist: pmdarima>=2.0.0
Provides-Extra: all
Requires-Dist: statsmodels>=0.14.0; extra == "all"
Requires-Dist: pmdarima>=2.0.0; extra == "all"
Dynamic: license-file

# Athena

**Automated EDA, ML readiness scoring, and data diagnostics — locally in your browser.**

```bash
pip install athena-eda
athena run
```

---

## What it does

Athena analyzes tabular datasets and surfaces everything that matters before you start modeling. Upload a CSV, pick a mode, and get a structured diagnostic report in under two minutes — no code required.

It covers the full pre-modeling workflow: data quality, feature analysis, ML readiness scoring, leakage detection, time series diagnostics, and multi-model comparison. When you're done, download a ready-to-run Python preprocessing script tailored to your dataset.

---

## Features

### Core EDA
- Column classification — numeric, categorical, identifiers, free text, entity keys
- Missing value analysis with row-level and column-level breakdown
- Outlier and skew analysis — per-feature skewness, kurtosis, log-transform recommendations
- Feature redundancy detection — correlation matrix analysis for highly redundant pairs

### ML Readiness
- **ML Readiness Score** — composite score across data health, trainability, and leakage risk
- **Leakage Detection** — flags identifier columns and suspiciously correlated features
- **Baseline Probe** — 3-fold cross-validated LightGBM baseline with learning curves and overfitting detection
- **Stability Analysis** — subsample stability checks across multiple runs
- **Clustering Diagnostics** — silhouette scoring across k=2 to 5 for unlabeled datasets
- **Anomaly Detection** — Isolation Forest with score distribution visualization

### Model Arena
- Auto-trains Random Forest, LightGBM, Logistic Regression (classification) or Ridge, Lasso, Random Forest, LightGBM (regression)
- Ranks all models by balanced accuracy or RMSE with training time and variance
- Only runs on labeled datasets with a specified target column

### Time Series
- Auto-detects datetime columns and forecast targets
- Stationarity testing — ADF and KPSS with differencing recommendations
- Seasonal decomposition with auto-detected period
- ACF and PACF with confidence bands
- Missing timestamp detection and coverage analysis
- Baseline models — Naive, Moving Average, Exponential Smoothing, Auto-ARIMA ranked by RMSE

### NLP
- Free-text column detection with vocabulary and lexical diversity analysis
- Text quality metrics — empty ratio, duplicate ratio, URL ratio
- Vectorization recommendations — TF-IDF vs embeddings
- Per-class word analysis for labeled datasets

### Drift Detection
- Upload train and test splits to get per-column KS-statistic drift scores
- New category detection and missing rate delta

### Outputs
- **Preprocessing script** — one-click export of a production-ready Python script tailored to your EDA results
- **HTML report** — full diagnostic report with charts, tables, and recommendations
- **Time series script** — pipeline script with load, stationarity check, decomposition, train/test split, and all baseline models

---

## Analysis Profiles

| Profile | Use case |
|---|---|
| Standard | General purpose defaults |
| Finance | Strict leakage detection, low anomaly tolerance |
| Healthcare | High missing value tolerance, strict cardinality thresholds |
| NLP / Text | Relaxed cardinality thresholds for text-heavy datasets |

---

## Installation

```bash
pip install athena-eda
```

Requires Python 3.9+. All dependencies including `statsmodels` and `pmdarima` for time series are included in the base install.

### Running locally

```bash
athena run
```

Opens your browser at `http://127.0.0.1:8000` automatically.

### Other commands

```bash
athena info      # show system and package info
athena doctor    # check all dependencies
athena version   # show current version
athena update    # update to latest version
```

---

## Stack

**Frontend** — React, TypeScript, Vite, Recharts

**Backend** — FastAPI, Python, LightGBM, scikit-learn, statsmodels, pmdarima, pandas

---

## License

MIT
