1"""
 2
 3# panelsplit: A tool for panel data analysis
 4
 5panelsplit is a Python package designed to facilitate time series cross-validation for panel (multi-entity) data. Whether you're in feature engineering, hyper-parameter tuning, or model estimation, panelsplit provides robust modules that make working with panel data both easier and more efficient.
 6
 7**Key Features:**
 8- **Panel data cross-validation:** Split up your panel dataset, respecting its temporal structure.
 9- **Data compatibility:** Provides a DataFrame-agnostic approach via narwhals, supporting libraries like polars, pandas, and numpy.
10- **Flexible pipelines:** Easily build pipelines that integrate with popular libraries such as scikit-learn and feature-engine.
11- **Parallel Processing:** Leverages parallel computing to speed up fitting and prediction tasks.
12
13---
14
15## Why choose panelsplit?
16
17panelsplit is built with the practical needs of data scientists working with panel data in mind:
18- **Robust & flexible:** Whether experimenting with models or deploying production pipelines, its modular design lets you focus on analysis rather than plumbing.
19- **User-friendly:** Clear API design and comprehensive documentation make it easy to integrate into your existing workflows.
20- **Efficient:** Parallel processing and tailored cross-validation ensure that your computations are both fast and accurate.
21
22Explore the modules in detail by clicking on the links below to see full documentation and examples.
23
24---
25## Modules
26
27
28### `panelsplit.cross_validation`
29- **PanelSplit class:** Automatically generates train/test splits while preserving temporal order and handling edge cases.
30- **Label generation:** Provides helper functions to create training and testing labels.
31- **Snapshot generation:** Generates snapshots of data in cases where transformations aren't comparable across time.
32
33### `panelsplit.application`
34- **Model fitting & prediction:** Fits models on each training split using cloned estimators and supports multiple prediction methods (e.g., `predict`, `predict_proba`).
35- **Parallel execution:** Leverages parallel processing for efficient handling of cross-validation splits.
36- **Data integrity:** Restores predictions to the original data order for consistency.
37
38### `panelsplit.pipeline`
39- **Sequential processing:** Chains multiple transformers and estimators into a streamlined pipeline, each with its own cross-validation approach.
40- **Dynamic method injection:** Automatically creates methods (like `predict` and `score`) based on the final estimator’s capabilities.
41- **Out-of-fold predictions:** Supports cross-validation based predictions with reassembled outputs.
42
43### `panelsplit.plot`
44- Visualize time series splits easily.
45"""
46
47__all__ = ["application", "cross_validation", "pipeline", "plot"]