Imodels: Interpretable ML Models

Understand the project

Practical overview

Use it at the modeling stage when interpretability, concise decision rules or visible tree structure are important alongside predictive performance.

Before you choose

The repository is public and identified with an MIT license. The documented installation method is `pip install imodels`.

The scikit-learn-compatible interface can fit workflows already organized around fit, predict and predict_proba.
The catalogue spans rule sets, rule lists and tree models rather than a single interpretability technique.
The printed tree example exposes feature thresholds and leaf values directly.
Some Bayesian rule-based models are documented as slow.
The repository lists Python 3.10–3.13 and NumPy >=2.0 in its installation badges.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up Imodels and train the documented HSTreeClassifierCV example on a sample dataset.

Install the package with the command below.
Documented command
```
pip install imodels
```
Import get_clean_dataset and HSTreeClassifierCV from imodels, plus train_test_split from scikit-learn.
Load the csi_pecarn_pred dataset and split it into training and test sets.
Initialize HSTreeClassifierCV(max_leaf_nodes=4), fit it with the training data and feature names, then call predict and predict_proba on the test data.
Print the fitted model to inspect its threshold-based tree representation.

A fitted interpretable tree model, discrete predictions, predicted probabilities and a readable printed tree.

The documented example uses a sample clinical dataset and a specific tree model; it does not document deployment or production validation.
The package documentation notes that some Bayesian models are slow.

Evidence and transparency

Source record details

Source availability: Active
Sources checked: 30 September 2026
Original website title: csinva/imodels · Official homepage title · retained in its source language