ComparaciĆ³n de algoritmos en Machine Learning aplicado en ventas de microempresas
Cargando...
Archivos
Fecha
2021
Profesor/a GuĆa
Facultad/escuela
Idioma
es
TĆtulo de la revista
ISSN de la revista
TĆtulo del volumen
Editor
Universidad AndrƩs Bello
Nombre de Curso
Licencia CC
Licencia CC
Resumen
El presente trabajo de tesis fue desarrollado, con el objetivo de modelar una herramienta
para microempresas capaz de mejorar la oportunidad de venta de productos en
licitaciones del sector pĆŗblico utilizando Machine Learning, en el proyecto trabajamos
sobre una empresa que sĆ³lo vende artĆculos deportivos por redes sociales, en ella, se
identificarƔ la problemƔtica y causas, se fijaran los objetivos y mƩtricas, luego de definir
el alcance se realiza un estudio de la utilizaciĆ³n de machine learning en la industria,
ademĆ”s de un anĆ”lisis teĆ³rico de los fundamentos y clasificaciĆ³n de algoritmos,
posteriormente una investigaciĆ³n de los avances en TI a nivel local y en la ciencia de
datos. Para construir la herramienta, es imprescindible contar con la informaciĆ³n
histĆ³rica, para lo cual se dedicĆ³ un capĆtulo dedicado a la obtenciĆ³n de los datos por la
extensiĆ³n y complejidad. A su vez, se implementĆ³ un ambiente de desarrollo para la
extracciĆ³n de los datos y set de pruebas, para desarrollar tanto los conectores que
permitan rescatar los datos pĆŗblicos desde Internet, como procesar la informaciĆ³n con
los distintos algoritmos de machine learning que serƔn evaluados.
Para el anƔlisis predictivo, se realizarƔn distintos experimentos en el lenguaje R,
utilizando como base de datos la informaciĆ³n recolectada, sometiendo a esta a los
distintos algoritmos de aprendizaje automƔtico supervisados que se utilizarƔn para
generar modelos de predicciĆ³n. Finalmente, los diferentes modelos se compararĆ”n para
seleccionar el de mejor rendimiento en nuestro modelo.
This thesis work was developed, with the aim of modeling a tool for micro-companies capable of improving the opportunity to sell products in public sector tenders using Machine Learning, in the project we work on a company that only sells sporting goods through social networks , in it the problem and causes will be identified, the objectives and metrics will be set, after defining the scope, a study of the use of machine learning in the industry is carried out, in addition to a theoretical analysis of the foundations and classification of algorithms, later An investigation of advances in IT at the local level and in data science. To build the tool, it is essential to have historical information, for which a chapter dedicated to obtaining the data due to the length and complexity was dedicated. At the same time, a development environment was implemented to extract the data and set of tests, to develop both the connectors that allow the rescue of public data from the Internet and process the information with the different machine learning algorithms that will be evaluated. For predictive analysis, different experiments will be carried out in the R language, using the collected information as a database, subjecting it to the different supervised machine learning algorithms that will be used to generate prediction models. Finally, the different models will be compared to select the one with the best performance in our model.
This thesis work was developed, with the aim of modeling a tool for micro-companies capable of improving the opportunity to sell products in public sector tenders using Machine Learning, in the project we work on a company that only sells sporting goods through social networks , in it the problem and causes will be identified, the objectives and metrics will be set, after defining the scope, a study of the use of machine learning in the industry is carried out, in addition to a theoretical analysis of the foundations and classification of algorithms, later An investigation of advances in IT at the local level and in data science. To build the tool, it is essential to have historical information, for which a chapter dedicated to obtaining the data due to the length and complexity was dedicated. At the same time, a development environment was implemented to extract the data and set of tests, to develop both the connectors that allow the rescue of public data from the Internet and process the information with the different machine learning algorithms that will be evaluated. For predictive analysis, different experiments will be carried out in the R language, using the collected information as a database, subjecting it to the different supervised machine learning algorithms that will be used to generate prediction models. Finally, the different models will be compared to select the one with the best performance in our model.
Notas
Tesis (MagĆster en GestiĆ³n de TecnologĆas de la InformaciĆ³n y Telecomunicaciones)
Palabras clave
Ropa Deportiva, ComercializaciĆ³n, Procesamiento ElectrĆ³nico de Datos, Aprendizaje de MĆ”quina