Scholarlink

<p>The exponential and dynamic growth of data underscores the need to efficiently execute Machine Learning (ML) projects to maximize their utility. However, ML methodologies have not kept pace with the rapid advances in data collection technology and artificial intelligence (AI). Notably, many methodologies developed over 30 years ago are still in use today. Given the advances in various fields of AI, there is an opportunity to analyze existing methodologies to enhance the effective application of ML algorithms. This study aims to provide an overview of ML methodologies introduced since 1989, highlighting their evolution and challenges, as well as the gaps that have been overlooked. We categorize these methodologies into three types: <i>research-oriented</i>, <i>industry-oriented</i>, and <i>agile environments-oriented</i>. Furthermore, we emphasize the importance of integrating Ethical, Legal, and Social Aspects (ELSA) into ML methodologies to ensure responsible and transparent AI development. We believe that this categorization, along with the incorporation of ELSA considerations, will assist practitioners in selecting the most suitable methodology for their projects. Additionally, we trust that the findings presented here will serve as a reflective foundation for the development and innovation of new ML methodologies, ensuring they meet the demands of modern data environments and contribute to advances in industrial and scientific development.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning methodologies: history and challenges

  • 期刊论文

  • 发表日期：

    2025年5月19日

查看全文
  • Cecilia-Irene Loeza-Mejía,
  • Eddy Sánchez-DelaCruz,
  • Irahan-Otoniel José-Guzmán

摘要

The exponential and dynamic growth of data underscores the need to efficiently execute Machine Learning (ML) projects to maximize their utility. However, ML methodologies have not kept pace with the rapid advances in data collection technology and artificial intelligence (AI). Notably, many methodologies developed over 30 years ago are still in use today. Given the advances in various fields of AI, there is an opportunity to analyze existing methodologies to enhance the effective application of ML algorithms. This study aims to provide an overview of ML methodologies introduced since 1989, highlighting their evolution and challenges, as well as the gaps that have been overlooked. We categorize these methodologies into three types: research-oriented, industry-oriented, and agile environments-oriented. Furthermore, we emphasize the importance of integrating Ethical, Legal, and Social Aspects (ELSA) into ML methodologies to ensure responsible and transparent AI development. We believe that this categorization, along with the incorporation of ELSA considerations, will assist practitioners in selecting the most suitable methodology for their projects. Additionally, we trust that the findings presented here will serve as a reflective foundation for the development and innovation of new ML methodologies, ensuring they meet the demands of modern data environments and contribute to advances in industrial and scientific development.