Una actualización del algoritmo de aprendizaje de reglas difusas de Wang y Mendel para problemas de clasificación con datos masivos

Laburpena

In recent decades, society has witnessed an unprecedented technological transformation. This period has been characterized by the widespread adoption of the Internet, the massive proliferation of mobile devices, and an astonishing increase in data generation. In this context, data analysis has emerged as one of the fastest-growing fields. Specifically, the analysis of massive data, part of what is known as Big Data, has become a fundamental approach to extract knowledge from human behavior and their environment. As a result, many organizations, both in the business and government sectors, have opted to employ these technologies to harness the immense potential of their data and extract valuable insights. However, this abundance of information presents a significant challenge. While this vast amount of data has the potential to significantly improve the accuracy of data mining algorithms, traditional approaches in this field are ill-equipped to handle the speed and volume requirements that Big Data demands. Therefore, it is necessary to develop new techniques to address these issues and enable effective analysis of this massive data. In this context, the present thesis focuses on the challenge of extracting meaningful knowledge from these vast volumes of data, with a primary emphasis on machine learning. This discipline, which is part of artificial intelligence, empowers machines to acquire knowledge directly from data and make autonomous decisions without human intervention. Machine learning relies on statistical techniques and algorithms designed to analyze data and reveal underlying patterns. Its applications are diverse, ranging from fraud detection to medical diagnosis. In this research, particular emphasis is placed on supervised learning for classification, a scenario in which labels are assigned to data to categorize them accurately and effectively. A rule-based learning model grounded in fuzzy logic takes center stage in this research. These models specialize in handling the uncertainty and ambiguity inherent in data. As a starting point, the Wang and Mendel algorithm (WM) is selected, recognized for its simplicity and efficiency when dealing with massive data, albeit accompanied by certain limitations in terms of accuracy and interpretability. The primary objective of this thesis is to significantly enhance the performance of the WM algorithm, especially when dealing with massive datasets.