Análisis de Rendimiento de Librerías de aprendizaje automático a través de lenguaje de programación Python para hacer predicciones, estimaciones y clasificadores

Abstract

This work presents a performance analysis of machine learning libraries through Python programming language. This analysis was done to demonstrate the advantages of this programming language in the field of data science. The basic information that has been analyzed with the machine learning libraries, are data from the National Institute of Statistics and Censuses with the theme marriages and divorces in Ecuador. To obtain a comparison balanced of libraries performance, the experimentation was done on the same set of training (80% Data) and testing (20% Data) in each of the libraries. This comparison was useful in order to determinate among other parameters, speed and accuracy, offered by some Python's machine learning libraries. In general, based on a number of parameters evaluation among them “Accuracy” (accuracy of the model) in the experiment of classification and “MSE” (mean square error) in the experiment of regression, the library Scikit-Learn obtained the best performance, on these two approaches.