Meta-Aprendizagem para seleção de técnicas de redução de dimensionalidade em problemas de Big Data

Abstract

Universities and industries generate vast amounts of data, often characterized by high dimensionality, which can negatively impact the performance of Machine Learning algorithms. Dimensionality reduction becomes a crucial solution to simplify these data without losing important information, allowing for more efficient analysis. However, manually selecting the most sui table dimensionality reduction algorithm for each dataset is a complex and time-consuming process. To automate this selection, this study proposes the development of a meta-learner capable of predicting which dimensionality reduction algorithm will be most efficient for a given dataset. This work focuses on the automated selection of dimensionality reduction algorithms in Big Data scenarios, utilizing meta-learning to identify patterns between datasets and the algorithms that yield the best results. The approach involves testing different dimensionality reduction techniques and, through the construction of comparative rankings, verifying whether the algorithm recommendation model can accurately predict the most suitable algorithm.