Uso de herramientas de Machine Learning para predecir dificultades económicas en empresas Pymes del sector alimenticio de productos a base de harina de la ciudad de Guayaquil a causa de la pandemia por Covid-19.
Fecha
2021
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad de Guayaquil. Facultad de Ciencias Matemáticas y Físicas. Carrera de Ingeniería en Sistemas Computacionales.
Resumen
Ante la crisis económica a nivel mundial desatada por el COVID-19, las PYMES del sector alimenticio se han visto afectadas, por ende, es necesario buscar soluciones que puedan conllevar a la reducción del impacto financiero a corto, mediano y largo plazo. El presente trabajo investigativo, se fundamenta en diseñar un modelo predictivo que sea capaz de, a través de algoritmos de aprendizaje supervisado de Machine Learning, tomar decisiones que ayuden a evaluar las dificultades económicas de empresas PYMES enfocadas al sector alimenticio. Bajo de este criterio, está como uno de los objetivos planteados el poder recopilar datos históricos que vayan acordes al comportamiento económico y financiero de una empresa durante el transcurso de la pandemia, de tal manera que se pueda elaborar un dataset y brindar un argumento ante los expertos del área financiera. Para ello, se cuenta con la metodología Knowledge Discovery in Databases KDD que consta de seis fases que giran en torno a la importación, muestreo, calidad, transformación, modelización, evaluación e implementación de los datos; teniendo en cuenta que la última fase será llevada a cabo a nivel de Python. Mediante el uso de librerías de Python como "sklearn" se obtuvo un porcentaje de precisión (96% Random Forest y 97% SVM), escogiendo entre ambos el último algoritmo puesto que presenta mayor sencillez, precisión y eficiencia para el manejo de recursos computacionales. De esta forma y habiendo obtenido una relación entre las variables mediante el coeficiente de correlación de Pearson, las hipótesis planteadas se cumplen, logrando con este proyecto beneficiar a todos aquellos que pertenezcan o estén involucrados con la gestión y manejo de recursos financieros, económicos o contables dentro de una PYMES.
Faced with the global economic crisis unleashed by COVID-19, SMEs in the food sector have been affected, therefore, it is necessary to find solutions that can lead to a reduction of the financial impact in the short, medium, and long term. The present investigative work is based on designing a predictive model that is capable of, through supervised learning algorithms of Machine Learning, making decisions that help to evaluate the economic difficulties of SMEs focused on the food sector. Under this criterion, one of the proposed objectives is to be able to collect historical data that is consistent with the economic and financial behavior of a company during the pandemic, in such a way that a dataset can be elaborated and provide an argument before the experts from the financial area. For this, the Knowledge Discovery in Databases KDD methodology is used, which consists of six phases that revolve around the import, sampling, quality, transformation, modeling, evaluation, and implementation of the data; keeping in mind that the last phase will be carried out at the Python level. Using Python libraries such as "sklearn" a percentage of precision was obtained (96% Random Forest and 97% SVM), choosing between them the last algorithm since it presents greater simplicity, precision, and efficiency for the management of computational resources. In this way and having obtained a relationship between the variables by means of the Pearson correlation coefficient, the hypotheses raised are fulfilled, achieving with this project benefit all those who belong to or are involved with the management and handling of financial, economic, or accounting resources within a SME.
Faced with the global economic crisis unleashed by COVID-19, SMEs in the food sector have been affected, therefore, it is necessary to find solutions that can lead to a reduction of the financial impact in the short, medium, and long term. The present investigative work is based on designing a predictive model that is capable of, through supervised learning algorithms of Machine Learning, making decisions that help to evaluate the economic difficulties of SMEs focused on the food sector. Under this criterion, one of the proposed objectives is to be able to collect historical data that is consistent with the economic and financial behavior of a company during the pandemic, in such a way that a dataset can be elaborated and provide an argument before the experts from the financial area. For this, the Knowledge Discovery in Databases KDD methodology is used, which consists of six phases that revolve around the import, sampling, quality, transformation, modeling, evaluation, and implementation of the data; keeping in mind that the last phase will be carried out at the Python level. Using Python libraries such as "sklearn" a percentage of precision was obtained (96% Random Forest and 97% SVM), choosing between them the last algorithm since it presents greater simplicity, precision, and efficiency for the management of computational resources. In this way and having obtained a relationship between the variables by means of the Pearson correlation coefficient, the hypotheses raised are fulfilled, achieving with this project benefit all those who belong to or are involved with the management and handling of financial, economic, or accounting resources within a SME.
Descripción
PDF
Palabras clave
Machine Learning, Covid-19, Random Forest, Svm, Dificultades Económicas, Pymes, Economic Difficulties, Smes