Machine Learning Regions of Reliability based on Sampling Distance Evaluation with Feature Decorrelation for Tabular Time Datasets

Author:

Askanazi Evan1,Grinberg Ilya1

Affiliation:

1. Bar Ilan University

Abstract

Abstract

Despite successful use in a wide variety of disciplines for data analysis and prediction, machine learning (ML) methods suffer from a lack of understanding of the reliability of predictions due to the lack of transparency and black-box nature of ML models. In materials science and other fields, typical ML model results include a significant number of low-quality predictions. This problem is known to be particularly acute for target systems which differ significantly from the data used for ML model training. However, to date, a general method for uncertainty quantification (UQ) of ML predictions has not been available. Focusing on the intuitive and computationally efficient similarity-based UQ, we show that a simple metric based on Euclidean feature space distance and sampling density together with the decorrelation of the features using Gram-Schmidt orthogonalization allows effective separation of the accurately predicted data points from data points with poor prediction accuracy. To demonstrate the generality of the method, we apply it to light GBM machine learning using a set of time series tabular data sets. We also show that this metric is a more effective UQ tool than the standard approach of using the average distance of k nearest neighbors (k = 1–10) in features space for similarity evaluation. The computational simplicity of this dataset combined with its applicability to time series datasets allows it to be readily used in numerous real world problems.

Publisher

Research Square Platform LLC

Reference40 articles.

1. Uncertainty Quantification Using Neural Networks for Molecular Property Prediction;Lior Hirschfeld Kyle;Journal of Chemical Information and Modeling,2020

2. Kevin Tran et al 2020 Mach. Learn.: Sci. Technol. 1 025006 (2022)

3. Uncertainty quantification for predictions of atomistic neural networks;Salazar L;Chem. Sci.,2022

4. Yuge Hu et al 2022 Mach. Learn.: Sci. Technol. 3 045028 (2022)

5. Uncertainty quantification in molecular simulations with dropout neural network potentials;Wen Mingjian;npj Computational Materials.,2020

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3