Miguel Angel Quiroz Martinez

Bio: Miguel Angel Quiroz Martinez is an academic researcher from Politecnica Salesiana University. The author has contributed to research in topics: TOPSIS & Deep learning. The author has an hindex of 2, co-authored 18 publications receiving 13 citations.

A Framework for Selecting Machine Learning Models Using TOPSIS

Abstract: In machine learning, it is common when multiple algorithms are applied to different data sets that are complex because of their accelerated growth, a decision problem arises, ie, how to select the algorithm with the best performance? This has generated the need to implement new information analysis techniques to support decision making The technique of multi-criteria decision making is used to select particular alternatives based on different criteria The objective of this article is to present some Machine Learning models applied to a data set in order to select the best alternative according to the criteria using the TOPSIS method The deductive method and the scanning research technique were applied to study a case study on the Wisconsin Breast Cancer dataset, which seeks to evaluate and compare the performance and effectiveness of machine learning models using the TOPSIS

Detección de Personal no Autorizado en el Departamento de TI Utilizando Redes Neuronales Convolucionales en Tiempo Real con Raspberry Pi 3 b

Abstract: espanolEl presente proyecto detalla la elaboracion de un sistema de reconocimiento facial y envio de alertas mediante correo electronico email. Para llevarlo a cabo, se utilizo la libreria TensorFlow, Numpy y OpenCV. Ademas, una Pi camara V2 y un Raspberry Pi 3 B+. Laslibrerias encargadas para alinear los rostros de las personas a detectar y luego usar estos rostros “alineados” para entrenar la red neuronal que se uso en la aplicacion son TensorFlow y Numpy. OpenCV para iniciar la captura de video y deteccion de objetos en tiempo real sobre el ambiente de un Raspberry Pi 3, encargado de procesar y almacenar la base de datos con las fotos de los rostros. Las alertas se envian mediante un correo electronico (email) por cada rostro no detectado y encendiendo un LED color rojo. Se implemento este sistema de reconocimiento facial con el objetivo de detectar rostros que no se encuentran dentro de la base de datos. El sistema de autenticacion se realizo en la Raspberry Pi 3, con la ayuda de las librerias TensorFlow, Numpy y OpenCV, los resultados obtenidos en este estudio fueron favorables luego de una evaluacion sobre la base de datos de 10 personas. Este sistema trabaja perfectamente con la base de datos, esta estudia, aprende los rostros captados y los compara con los almacenados en la base de datos; y alertara cuando detecte un rostro no encontrado. EnglishThis project details the development of a facial recognition system and the sending of alerts by email email. To carry it out, the TensorFlow, Numpy and OpenCV library was used. Also, a Pi V2 camera and a Raspberry Pi 3 B+. the libraries in charge of aligning the faces of the people to be detected and then using these “aligned” faces to train the neural network that was used in the application are TensorFlow and Numpy. OpenCV to initiate the capture of video and detectionof objects in real time on the environment of a Raspberry Pi 3, in charge of processing and storing the database with the photos of the faces. Alerts are sent by email (email) for each undetected face and lighting a red LED. This facial recognition systemwas implemented with the aim of detecting faces that are not found in the database. The authentication system was carried out on the Raspberry Pi 3, with the help of the TensorFlow, Numpy and OpenCV libraries, the results obtained in this study were favorable after an evaluation based on the data of 10 people. This system works perfectly with the database, it studies, learns the captured faces and compares them with those stored in the database; and will alert when it detects a face not found

Architectural Design for the Implementation of Learning Analytics: Case Study at Salesian Polytechnic University

Abstract: Learning analytics consist of measuring, capturing, and analyzing student data. Currently, the use of learning analytics in Universities constitutes a world trend that helps to make better academic decisions in the Salesian Polytechnic University with the purpose of decrease the dropout rates of the career. The institution developed tutoring processes in which help is provided to students in the teaching-learning process. The problem is that attendance at these tutorials is optional. There is a proliferation of concern about whether the students who most need it the most are the ones who attend these tutorials. Consequently, the University tries to identify the students that are most likely to drop out of their careers before it happens and offer them the help they need through these tutorials. Today, student’s data are growing and are indifferent and scattered databases. This article proposes an architectural design for the implementation of learning analytics that seeks to strengthen students to achieve their personal and academic goals. The evaluation of the architecture was based on the LWPM operator and shows a high evaluation, according to the experts, highlighting the maintainability and reliability of this architecture.

A Framework for Selecting Classification Models in the Intruder Detection System Using TOPSIS

Abstract: As the network has expanded considerably, security mechanisms are a key issue in networks. Intrusive activities, such as unauthorized access and data manipulation, are increasing. Therefore, the role of the Network Intrusion Detection System (NIDS) in monitoring network traffic for activity and determining whether an intrusion has occurred is very important. The performance of an IDS depends on the selection of the classification model and training data, however, many classifiers generate similar results when measuring performance. The technique of order of preference for similarity to the ideal solution (TOPSIS) is used to select one or more alternatives based on the criteria. The main objective is to present some classification models used in a data set to select the best alternative according to the performance criteria using the TOPSIS method. The deductive method and selection research technique were applied to study the NSL-KDD.

Selection of LPWAN Technology for the Adoption and Efficient Use of the IoT in the Rural Areas of the Province of Guayas Using AHP Method

Abstract: Ecuador, being a developing country, depends on the search and implementation of new technologies that help improve the productive matrix, this is the case of the agricultural sector, driven by the growth of the Internet market Internet of Things (IoT) Low-power wide-area network (LPWAN) has allowed its expansion in underdeveloped countries Sigfox, LoRa, and NB-IoT are the three leading LPWAN technologies that compete for large-scale IoT implementation This document provides a comparative study of these technologies, focused on the livestock sector Basic criteria are identified to facilitate decision making when using IoT technology for the development of livestock in rural areas of Guayas province Through the analytical hierarchical process(AHP), the type of technologies to be implemented is ranked and selected

Adapted fusion schemes for multimodal biometric authentication (Esquemas adaptados de fusión para autenticación biométrica multimodal)

Abstract: Se denomina reconocimiento biometrico al proceso que permite asociar una identidad con un individuo de forma automatica, mediante el uso de alguna caracteristica personal que le sea inherente. El objetivo en la verificacion biometrica es decidir entre dos clases, cliente o impostor. Para tal finalidad hay una serie de modalidades fisiologicas que pueden ser consideradas como: la huella dactilar, el iris, la cara, la geometria de los dedos y/o la mano, o la huella palmar, la voz, la escritura y la firma manuscrita, o el modo de andar (marcha), ... Esta Tesis se centra en la combinacion de varios rasgos biometricos para superar algunas de las limitaciones de rasgos individuales, en lo que se conoce como un sistema biometrico multimodal. La mayoria de estrategias para la fusion multimodal existentes en la literatura se basan en la fusion de las puntuaciones proporcionadas por los sistemas individuales. En la mayor parte de los casos se asume que dichos esquemas de combinacion se mantienen fijos durante la etapa de verificacion. Esta Tesis se centra en el estudio de esquemas adaptados de fusion durante la etapa de verificacion y se plantean nuevas arquitecturas de fusion para autenticacion biometrica adaptadas tanto a los diferentes usuarios registrados en el sistema, como a la calidad de las senales biometricas de entrada; resultando en esquemas adaptados de fusion para autenticacion biometrica multimodal. Estos esquemas encuentran aplicacion asimismo en otros escenarios que combinan diferentes fuentes de informacion biometrica y no necesariamente diferentes rasgos, como son la fusion multi-sensor, multi-algoritmo, multi-unidad, y de unidades repetidas. El termino 'adaptado' en esta Tesis se refiere a esquemas entrenados utilizando informacion general, por ejemplo un conjunto de usuarios de referencia, y ajustados al considerar informacion especifica bien del usuario que se trate o bien de la calidad de la muestra biometrica de

Redes neuronales artificiales

Abstract: Una RNA (Red Neuronal Artificial) es un modelo matemático inspirado en el comportamiento biológico de las neuronas y en la estructura del cerebro. Esta también puede ser vista como un sistema inteligente que lleva a cabo tareas de manera distinta a como lo hacen las computadoras actuales. Si bien estas últimas son muy rápidas en el procesamiento de la información, existen tareas muy complejas, como el reconocimiento y clasificación de patrones, que demandan demasiado tiempo y esfuerzo aún en las computadoras más potentes de la actualidad, pero que el cerebro humano es más apto para resolverlas, muchas veces sin aparente esfuerzo (considere el lector como ejemplo el reconocimiento de un rostro familiar entre una multitud de otros rostros). El cerebro puede considerarse un sistema altamente complejo. Su unidad básica, la neurona, está masivamente distribuida con conexiones entre ellas (se calcula que hay aproximadamente 10 billones de neuronas en la corteza cerebral y 60 trillones de conexiones neuronales). Si bien hay distintos tipos de neuronas biológicas, en la figura 1 se muestra un esquema simplificado de un tipo particular que es muy común. Vemos que la misma está compuesta por:

Neutrosophic-Simplified-Topsis. Multi-Criteria Decision-Making Using Combined Simplified-Topsis Method and Neutrosophics

Abstract: In this paper, (1) one simplified the standard TOPSIS to new Multi-Criteria Decision-Making (MCDM) called Simplified-TOPSIS. Simplified-TOPSIS gives the same results and simplifies the calculation of the classical TOPSIS. An example is presented distinctions between Simplified-TOPSIS and classical TOPSIS are underlined. (2) extend the new Simplified-TOPSIS method to Neutrosophic-simplified-TOPSIS using single valued Neutrosophic information. An example showing the interest of Neutrosophic-simplified-TOPSIS to manipulate the uncertainty linked to information presented in Multi-Criteria Decision-Making.

Profiles of Human Trafficking Violence in Regions of Ecuador

Abstract: People are often vulnerable to become victims of violence, including human trafficking. In Ecuador, considerable crimes of this nature are reported each year. Knowing and generating protection mechanisms against this type of crime constitutes a vitally important task to preserve the integrity of people. This research proposes a solution to the problem described from the development of a method to determine profiles of violence in people from regions of Ecuador. The proposed method bases its operation on the set of neutrosophic numbers to represent the uncertainty.

Implementación de un modelo de reconocimiento visual para mejorar el modelo adaptativo en niños con daltonismo usando un robot Nao para niños vulnerables en una ciudad inteligente

Abstract: Using a Humanoid robot Nao, we propose support to children whom have a specific color deficiency named “Color Blindness”, which prevents this minority group to be alert of possible visual warning in games, parks and zoos. The relevance of our study lies in the support in situations of danger on the part of children with color blindness, identifying threats with specific colors and to