Algorithmic bias reduction strategies in machine learning
Department of Electrical and Computer Engineering, Hampton University, Hampton, Virginia, United States.
Research Article
Open Access Research Journal of Engineering and Technology, 2026, 10(02), 170-177.
Article DOI: 10.53022/oarjet.2026.10.2.0046
Publication history:
Received on 24 April 2026; revised on 01 June 2026; accepted on 03 June 2026
Abstract:
As Artificial Intelligence (AI) and Machine Learning (ML) systems are increasingly integrated into critical societal domains such as finance, healthcare, and autonomous infrastructure, the prevalence of algorithmic bias has emerged as a paramount global concern. This research addresses the lack of a standardized, comprehensive understanding of how systematic deviations enter the AI lifecycle, leading to social harm, eroded public trust, and significant legal risks. Algorithmic bias is defined as systematic, non-random errors in machine learning algorithms that generate outcomes disproportionately favoring or disfavoring only one side. At the societal level, Fink's Seven-Phase Issue-Development Process is utilized to monitor the evolution of algorithmic failures—such as gender disparities in search results and racial inaccuracies in computer vision—before they escalate into public crises. At the technical level, the research evaluates the efficacy of selecting best-suited estimators and optimizers. Empirical comparisons conducted via Python-based benchmarking demonstrate that architectures such as Convolutional Long Short-Term Memory (CLSTM) and Extra Trees (ET) significantly reduce systematic deviation in imbalanced datasets compared to sub-optimal models like Multinomial Naive Bayes (MNB) or Gradient Boosted Regression Trees (GBRT). The findings underscore that effective bias reduction requires a dual-track strategy which includes aligning technical model selection with the best data system and implementing ethical auditing throughout the development pipeline. This research provides a robust framework for training data best practices and the need to use the best estimators, optimizers, and regressors in machine learning.
Keywords:
Algorithm Bias; Machine Learning; Taxonomy Development; Framework Integration; Technical Benchmarking; Systematic Mitigation; Optimizers; Estimators; Decision Tree Regressor; Linear Regression
Full text article in PDF:
Copyright information:
Copyright © 2026 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution Liscense 4.0