K-Nearest neighbor in a classification and prediction application in the Judicial Power of Peru

• Autor: Quezada Lucio, Nel
• Formato: artículo
• Fecha de Publicación: 2018
• Institución: Universidad Nacional Mayor de San Marcos
• Repositorio: Revistas - Universidad Nacional Mayor de San Marcos
• Lenguaje: español
• OAI Identifier: oai:revistasinvestigacion.unmsm.edu.pe:article/15077
• Enlace del recurso: https://revistasinvestigacion.unmsm.edu.pe/index.php/matema/article/view/15077
• Nivel de acceso: acceso abierto
• Materia: clasificación supervisada y no supervisada; k-vecinos más próximos; clasificación no paramétrica; partición; validación cruzada aleatoria.; supervised and unsupervised classification; nearest k-neighbors; nonparametric classification; partition; random cross-validation

Abstract: This article summarizes the main contributions of the thesis with the title “K-Nearest neighbor in a classification and prediction application in the Judicial Power of Peru". In this thesis a model is constructed using the method of the nearest k-neighbors that allows classifying and predicting the Superior Courts of Justice of Peru. Through a descriptive data analysis, the Lima Court is excluded from the study. With the remaining 30 Superior Courts, a three-group model based on unsupervised classification is generated, for which the Euclidean distance matrix that originates the classification tree is deduced. The classification model of three nearest neighbors is constructed, with partition and random cross-validation folds, which indicates; the predictor space model, the quadratic error or error index that validates the op-timal value of k = 3 neighbors, the model error and the global forecast index that measure the accuracy or accuracy of the model found, importance of the predictor, maps of quadrants and table of neighbors.