Procesamiento de bases de datos escolares por medio de redes neuronales artificiales

Abstract

El estudio de bases de datos escolares es un área que ha sido poco estudiada y cuestionada desde el punto de vista de la minería de datos o la inteligencia artificial, actualmente, existen algunos trabajos que muestran su procesamiento mediante algoritmos de aprendizaje automático o “inteligentes”, sin embargo, no se detienen a analizar la pertinencia de procesar datos cualitativos como si fueran cuantitativos. En este trabajo, se estudia este problema con el uso de tres modelos de red neuronal. Los resultados evidencian la capacidad de estos modelos para clasificar con un alto grado de acierto tendencias en los estudiantes, utilizando principalmente datos cualitativos.

Neural Network for Vocational Guidance Based on The Applicant's Profile For Admission To A Study Program

“…Note: own elaboration Figure 2 shows the neural network constructed in which 40 data related to the competencies of the entry profile classified as follows 3 axón's (X1, X2, X3), each incoming axon generated 18 synapses or weights (W11 (1) , …, W21 (1) , …, W31 (1) , …, W36 (1) ), both are calculated by means of the aggregation function (Z1 (2) , Z2 (2) , Z3 (2) , Z4 (2) , Z5 (2) , Z6 (2) ) complying with: Z (2) = X*W (1) and the construction of the activation function a (2) = f(Z (2) ) represented by the sigmoid function for the hidden layer; the protruding axon represented by "y" is the operation between the activation functions (a1 (2) , a2 (2) , a3 (2) , a4 (2) , a5 (2) , a6 (2) ) and their respective weights (W11 (2) , W21 (2) , W31 (2) , W41 (2) , W51 (2) , W61 (2) ) constructing the function Z (3) = a (2) *W (2) and activation function y = f(Z (3) ) the function used is the sigmoid. The results found are not completely reliable since the percentage of reliability is not high, so we have resorted to identifying the cost function with the mean squared error J = ½*e 2 and the gradient descent algorithm with the derivative of the function and to ensure high reliability the second derivative of the function was used (LeCun et al, 2012).…”

“…Artificial Neural Network. It represents the nervous system of the human brain through mathematical algorithms capable of solving complex problems, (Haykin, 2009) cited by García et al (2020) consider that an ANN is a "... mathematical model that tries to emulate the biological neuronal systems of the human being in the processing of information" (p. 443). See Figure 1.…”