tidy.outliers

tidymodels

code

package

Author

Bruno Testaguzza Carlin

Published

January 31, 2023

What is tidy.outliers?

tidy.outliers is a pet project of mine born out of a project when I saw a coworker manually implement an outlier detection algorithm and wondered if tidy.models had an easy way to do it.

I knew that scikit had up to 5 established ways of detecting outliers and even a great page talking about outlier removal

This situation left me wondering why no one had written something similar for the incredible tidymodels ecosystem. So I decided to do it myself.

What v 0.2.0 adds?

Univariate methods

I have added a method for users to pass a function to outlier steps, so now if you can have your custom univariate based rules to score outliers, using

step_outliers_univariate

h2o integration

With the new step step_outliers_h2o.extendedIsolationForest, you can read more about the function on Extended Isolation Forest .

outForest method

The new step_outliers_outForest uses the main function from the package, you can read more about it here outForest.

Improvements to CI/CD

The package reached 100% test coverage, and it currently passes four out of five setups for a possible cran release. I even posted my first covering it.

I also upgraded the package to the v2 action framework of the rstudio team, and you can read more about it actions

---
title: "tidy.outliers"
author: "Bruno Testaguzza Carlin"
toc: true
df-print: paged
code-tools: true
code-fold: show
code-link: true
categories: [tidymodels,code,package]
image: https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSDvE-FcZtiyIvEPBjf_UGXqeEOXh-iVx2K3g&usqp=CAU
date: 2023-01-31
---

# What is tidy.outliers?

tidy.outliers is a pet project of mine born out of a project when I saw a coworker manually implement an outlier detection algorithm and wondered if tidy.models had an easy way to do it.

I knew that scikit had up to 5 established ways of detecting outliers and even a [great page talking about outlier removal](https://scikit-learn.org/stable/modules/outlier_detection.html "great page talking about outlier removal")

This situation left me wondering why no one had written something similar for the incredible tidymodels ecosystem. So I decided to do it myself.

# What v 0.2.0 adds?

## Univariate methods

I have added a method for users to pass a function to outlier steps, so now if you can have your custom univariate based rules to score outliers, using

[step_outliers_univariate](https://brunocarlin.github.io/tidy.outliers/reference/step_outliers_univariate.html "step_outliers_univariate")

## h2o integration

With the new step [step_outliers_h2o.extendedIsolationForest](https://brunocarlin.github.io/tidy.outliers/reference/step_outliers_h2o.extendedIsolationForest.html "step_outliers_h2o.extendedIsolationForest"), you can read more about the function on [Extended Isolation Forest](https://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/eif.html "Extended Isolation Forest") .

## outForest method

The new [step_outliers_outForest](https://brunocarlin.github.io/tidy.outliers/reference/step_outliers_outForest.htm "step_outliers_outForest") uses the main function from the package, you can read more about it here [outForest](https://github.com/mayer79/outForest "outForest").

# Improvements to CI/CD

The package reached 100% test coverage, and it currently passes four out of five setups for a possible cran release. I even posted my first covering it.

[My first video](https://www.youtube.com/watch?v=U_Ahxcq_6Sk "My first video")

I also upgraded the package to the v2 action framework of the rstudio team, and you can read more about it [actions](https://github.com/r-lib/actions "actions")