Introducing HoNCAML: Holistic No-Code Auto Machine Learning

Luca Piras, Joan Albert Erráez Castelltort, Jordi Casals Grifell, Xavier de Juan Pulido, Cirus Iniesta, Marina Rosell Murillo, Cristina Soler Arenys

Keywords: automl, library, no-code, open-source

TL;DR: We present a new open-source no-code AutoML tool, implemented as a Python library

Abstract: In recent years, Machine Learning (ML) has been changing the landscape of many industries, demanding companies to incorporate ML solutions to stay competitive. In response to this imperative, and with the aim of making this technology more accessible, the emergence of "no-code" AutoML assumes critical significance. This paper introduces HoNCAML (Holistic No-Code Auto Machine Learning), a new AutoML library designed to provide an extensive and user-friendly resource accommodating individuals with varying degrees of coding proficiency and diverse levels of ML expertise, inclusive of non-programmers. The no-code principles are implemented through a versatile interface offering distinct modalities tailored to the proficiency of the end user. The efficacy of HoNCAML is comprehensively assessed through qualitative comparisons with analogous libraries, as well as quantitative performance benchmarks on several public datasets. The results from our experiments affirm that HoNCAML not only stands as an accessible and versatile tool, but also ensures competitive performance across a spectrum of ML tasks.

Submission Checklist: Yes

Broader Impact Statement: Yes

Paper Availability And License: Yes

Code Of Conduct: Yes

Optional Meta-Data For Green-AutoML: This blue field is just for structuring purposes and cannot be filled.

CPU Hours: 30

Community Implementations: https://github.com/Data-Science-Eurecat/HoNCAML

Submission Number: 11