Systematic Review of The Challenges in Implementing Data Segmentation with Machine Learning

Abstract

This study identifies the main challenges in implementing data segmentation using machine learning techniques. A systematic literature review was carried out using the PICO methodology and the PRISMA framework, which allowed the selection of 44 relevant articles. The predominant methods include Deep Learning techniques, Ensemble Learning and traditional classification approaches, applied in domains such as telecommunications, health and cybersecurity. Among the highlighted challenges are the high complexity of the data, the presence of noise, the inconsistent quality of the records and the difficulty in integrating heterogeneous sources. Despite the progress made, limitations persist in terms of scalability and the absence of standardized methodological frameworks. This study is very useful for researchers oriented to improve the precision and efficiency in data segmentation in environments with large volumes of information. The development of adaptive methodologies and the establishment of standards that facilitate the transfer of knowledge between sectors are proposed.