IntervalScaleClassification_Experimentation: Performance metrics for ordinal and interval scale classification

Rosario Delgado* (Programador), Giulia Binotto (Programador)

*Autor corresponent d’aquest treball

Producció científica: Formats no textualsProgramariRecerca

Resum

In the field of supervised machine learning, the precise evaluation of classification models stands as a fundamental pursuit. This necessitates the utilization of robust performance metrics. This repository focuses on the real-world application of the interval-scale measures introduced in the paper

"Adapting Ordinal Performance Metrics for Interval Scale: Length Matters" by G. Binotto and R. Delgado (preprint, 2024)

for evaluation and as metrics/error functions for hyper-parameter tuning in classification (experiments in Section 5 of the paper). We consider three real-world datasets:

Facial age dataset (https://www.kaggle.com/datasets/frabbisw/facial-age)
Abalone dataset (https://archive.ics.uci.edu/dataset/1/abalone)
Parkinson dataset (https://archive.ics.uci.edu/dataset/189/parkinson+telemonitoring)
It uses the content in https://github.com/giuliabinotto/ IntervalScaleClassification, which correspond to Section 4 fot the same paper, where scripts facilitate the computation of two ordinal metrics, Mean Absolute Error (MAE) and Total Cost (TC), alongside their interval scale counterparts introduced in the paper, with a specific section designed to address scenarios in which the rightmost interval is unbounded
Idioma originalAnglès
Format de la publicacióInternet
Estat de la publicacióPublicada - 2024

Fingerprint

Navegar pels temes de recerca de 'IntervalScaleClassification_Experimentation: Performance metrics for ordinal and interval scale classification'. Junts formen un fingerprint únic.

Com citar-ho