Computational Learning Models of Scikit-Learn for Automatic People Identification Integrated in a GUI
Carlos Vicente Niño Rondón, Byron Medina Delgado, Sergio Alexander Castro Casadiego, Jorge Gómez Rojas · 2022
This paper presents the analysis and weighting of the computational learning methods of classifiers of logistic regression, perceptron, Ridge and Passive-Aggressive from the Scikit-Learn library for a GUI-integrated automatic person identification system. Datasets for 10 classes are created using optimized cascade classifier techniques and the performance of the classifiers are trained and compared at hardware and software levels. The best performance was presented by the Ridge classifier together with the logistic regression classifier with 98.5 % in accuracy, being these also the best performers in accuracy per class with 98.52 %. At the hardware level, the Ridge classifier had the best performance, requiring only 6.9 % of CPU, 0.3166 seconds for training, and 216.2 MB of memory. The method and classifiers considered are efficient in a way that allows their replicability in access control systems on processors, microprocessors and embedded systems, and are shown to be an alternative to conventional face recognition detection methods based on deep learning.