Work

XNB: A package for Class-Specific Naive-Bayes classifier

Year: 2025

Type: software-paper

Abstract: The Explainable Class-Specific Naive Bayes ( XNB ) package is a novel software tool for classification, specifically developed for high-dimensional data scenarios where interpretability is essential. XNB enhances the traditional Naive Bayes model through two core innovations. First, it replaces the restrictive Gaussian assumption with Kernel Density Estimation (KDE), enabling more flexible and accurate modeling of complex, non-Gaussian distributions. Second, it incorporates a class-specific feature selection strategy, which identifies distinct subsets of relevant variables associated with each class. This selection mechanism improves interpretability and reduces both dimensionality and feature redundancy. Empirical evaluations on genomic datasets show that XNB achieves competitive classification performance while using dramatically fewer features. The combination of accurate density estimation and class-aware, sparse feature selection results in a transparent classification framework. This makes the XNB package particularly valuable in domains such as biomedicine, where software tools that combine precision and explainability are critically needed.

Source: SoftwareX

DOI: 10.1016/j.softx.2025.102347

Authors: Jesús S. Aguilar–Ruiz, Cayetano Romero–Vargas

Institutions: Universidad Pablo de Olavide, Indra (Spain)

Language: English

Topic: Explainable Artificial Intelligence (XAI)

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $1,200

Funder: Ministerio de Ciencia, Innovación y Universidades