Benchmarking Methods for Pointwise Reliability

The growing interest in machine learning in a critical domain like healthcare emphasizes the need for reliable predictions, as decisions based on these outputs can have significant consequences. This study benchmarks methods for assessing pointwise reliability, focusing on data-driven techniques bas...

Full description

Saved in:
Bibliographic Details
Main Authors: Cláudio Correia, Simão Paredes, Teresa Rocha, Jorge Henriques, Jorge Bernardino
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Information
Subjects:
Online Access:https://www.mdpi.com/2078-2489/16/4/327
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850181009481400320
author Cláudio Correia
Simão Paredes
Teresa Rocha
Jorge Henriques
Jorge Bernardino
author_facet Cláudio Correia
Simão Paredes
Teresa Rocha
Jorge Henriques
Jorge Bernardino
author_sort Cláudio Correia
collection DOAJ
description The growing interest in machine learning in a critical domain like healthcare emphasizes the need for reliable predictions, as decisions based on these outputs can have significant consequences. This study benchmarks methods for assessing pointwise reliability, focusing on data-driven techniques based on the density principle and the local fit principle. These methods evaluate the reliability of individual predictions by analyzing their similarity to training data and evaluating the performance of the model in local regions. Aiming to establish a standardized comparison, the study introduces a benchmark framework that combines error rate evaluations across reliability intervals with t-distributed Stochastic Neighbor Embedding visualizations to further validate the results. The results demonstrate that methods combining density and local fit principles generally outperform those relying on a single principle, achieving lower error rates for high-reliability predictions. Furthermore, the study identifies challenges such as the adjustment of method parameters and clustering limitations and provides insight into their impact on reliability assessments.
format Article
id doaj-art-e089fe7ed6944b6785f786e2f75151d2
institution OA Journals
issn 2078-2489
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Information
spelling doaj-art-e089fe7ed6944b6785f786e2f75151d22025-08-20T02:17:59ZengMDPI AGInformation2078-24892025-04-0116432710.3390/info16040327Benchmarking Methods for Pointwise ReliabilityCláudio Correia0Simão Paredes1Teresa Rocha2Jorge Henriques3Jorge Bernardino4Coimbra Institute of Engineering, Polytechnic University of Coimbra, Rua da Misericórdia, Lagar dos Cortiços, S. Martinho do Bispo, 3045-093 Coimbra, PortugalCoimbra Institute of Engineering, Polytechnic University of Coimbra, Rua da Misericórdia, Lagar dos Cortiços, S. Martinho do Bispo, 3045-093 Coimbra, PortugalCoimbra Institute of Engineering, Polytechnic University of Coimbra, Rua da Misericórdia, Lagar dos Cortiços, S. Martinho do Bispo, 3045-093 Coimbra, PortugalCISUC, Center for Informatics and Systems of University of Coimbra, University of Coimbra, Pólo II, 3030-290 Coimbra, PortugalCoimbra Institute of Engineering, Polytechnic University of Coimbra, Rua da Misericórdia, Lagar dos Cortiços, S. Martinho do Bispo, 3045-093 Coimbra, PortugalThe growing interest in machine learning in a critical domain like healthcare emphasizes the need for reliable predictions, as decisions based on these outputs can have significant consequences. This study benchmarks methods for assessing pointwise reliability, focusing on data-driven techniques based on the density principle and the local fit principle. These methods evaluate the reliability of individual predictions by analyzing their similarity to training data and evaluating the performance of the model in local regions. Aiming to establish a standardized comparison, the study introduces a benchmark framework that combines error rate evaluations across reliability intervals with t-distributed Stochastic Neighbor Embedding visualizations to further validate the results. The results demonstrate that methods combining density and local fit principles generally outperform those relying on a single principle, achieving lower error rates for high-reliability predictions. Furthermore, the study identifies challenges such as the adjustment of method parameters and clustering limitations and provides insight into their impact on reliability assessments.https://www.mdpi.com/2078-2489/16/4/327benchmarkdensitylocal fitmachine learningpointwise reliability
spellingShingle Cláudio Correia
Simão Paredes
Teresa Rocha
Jorge Henriques
Jorge Bernardino
Benchmarking Methods for Pointwise Reliability
Information
benchmark
density
local fit
machine learning
pointwise reliability
title Benchmarking Methods for Pointwise Reliability
title_full Benchmarking Methods for Pointwise Reliability
title_fullStr Benchmarking Methods for Pointwise Reliability
title_full_unstemmed Benchmarking Methods for Pointwise Reliability
title_short Benchmarking Methods for Pointwise Reliability
title_sort benchmarking methods for pointwise reliability
topic benchmark
density
local fit
machine learning
pointwise reliability
url https://www.mdpi.com/2078-2489/16/4/327
work_keys_str_mv AT claudiocorreia benchmarkingmethodsforpointwisereliability
AT simaoparedes benchmarkingmethodsforpointwisereliability
AT teresarocha benchmarkingmethodsforpointwisereliability
AT jorgehenriques benchmarkingmethodsforpointwisereliability
AT jorgebernardino benchmarkingmethodsforpointwisereliability