Influência da normalização no desempenho de classificadores monolíticos e combinados

Abstract

Classification is a process that uses data and algorithms to categorize objects based on specific characteristics. In statistics there are different approaches to classification, such as neural networks, ensemble methods, statistical models, support vector machines (SVM), among others. Therefore, it is important to choose an approach based on the specific characteristics of the problem and the data. This work aims to analyze whether the application of of variable scalling techniques influences the improvement of monolithic and combined classifiers. For this purpose, scaling techniques (normalization) were used and the influence on the performance of the classifiers was evaluated by measures, such as F1, accuracy, specificity and sensitivity. The study also includes a statistical definition of the concepts used and a methodological review, in addition to the data preprocessing phase.