Penalized accuracy for evaluating binary classifier and Ising models†
Donghyuk Lee1 · Jongmin Lee2 · Minwoo Kim3
Department of Statistics, Pusan National University
Correspondence to: † This work was supported by a 2-Year Research Grant of Pusan National University
1 Assistant professor, Department of Statistics, Pusan National University, Busan 46241, Korea.
2 Assistant professor, Department of Statistics, Pusan National University, Busan 46241, Korea.
3 Corresponding author: Assistant professor, Department of Statistics, Pusan National University, Busan 46241, Korea. E-mail: mwkim@pusan.ac.kr
1 Assistant professor, Department of Statistics, Pusan National University, Busan 46241, Korea.
2 Assistant professor, Department of Statistics, Pusan National University, Busan 46241, Korea.
3 Corresponding author: Assistant professor, Department of Statistics, Pusan National University, Busan 46241, Korea. E-mail: mwkim@pusan.ac.kr
Received September 19, 2024; Revised October 17, 2024; Accepted October 20, 2024.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
- Abstract
- Binary classification is a field of statistics that refers to the process of classifying given data into positive or negative categories. Logistic regression models, support vector machines, and neural networks are representative examples of binary classification models. A popular metric for evaluating the performance of binary classification models is an accuracy. An accuracy represents the proportion of true positives and true negatives among all predictions made by the classifier, essentially indicating the rate of correct predictions. While the accuracy is advantageous due to its intuitive understanding and is widely used in statistics, machine learning, and computer science, it has the drawback of not accounting for the effects of false positives and false negatives. This study presents specific examples where accuracy leads to distorted interpretations and defines a penalized accuracy that can be modified to suit specific purposes. Additionally, we proves the relationship between the maximum likelihood estimation of an Ising model and penalized accuracy, demonstrating the potential for more extended research.
- Keywords : Binary classification, ising model, maximum likelihood estimation, penalized accuracy
Full Text(PDF) Free
