Uso de algoritmos de Machine Learning para analizar los datos de energía eléctrica facturada en la Ciudad de Buenos Aires durante el período 2010–2021
Files
Date
Authors
Subject
Billed energy
Machine Learning
Linear Regression
Buenos Aires
Seasonality
energía facturada
Machine Learning
Regresión Lineal
Buenos Aires
estacionalidad
Machine Learning
Linear Regression
Buenos Aires
Seasonality
energía facturada
Machine Learning
Regresión Lineal
Buenos Aires
estacionalidad
Language:
Journal Title
Journal ISSN
Volume Title
Publisher
Instituto Tecnológico de Santo Domingo (INTEC)
Con las capacidades actuales de las computadoras, los algoritmos de Machine Learning se implementan con facilidad en distintas áreas de interés. En el presente estudio se utilizan métodos de Machine Learning para analizar los datos de energía eléctrica facturada mensualmente en la Ciudad de Buenos Aires, durante el período 2010-2021. Los objetivos son: determinar patrones en los datos utilizando el algoritmo K-Means y determinar las variables que más impactan la energía facturada total a través del uso del algoritmo de Regresión Lineal. Como técnica de reducción de la dimensionalidad se utilizó el análisis de componentes principales. La investigación fue de tipo cuantitativa-explicativa, utilizando los datos de la Dirección General de Estadística y Censos de Buenos Aires, los cuales fueron analizados y preprocesados antes de la aplicación de los algoritmos; para generar los modelos se toma el 75 % de los datos para entrenamiento y 25 % para la evaluación del modelo obtenido. Para el modelo de agrupamiento K-Means se determinó el K óptimo a través del método del codo, y se obtuvo que los datos de energía facturada total presentan una estacionalidad mensual. Para el modelo de Regresión Lineal se utilizaron las métricas R2, RMSE y MAE, y se obtuvo que las energías facturadas residencial, comercial e industrial, más el número de usuarios residenciales, son las variables que mayor impacto tienen sobre la energía eléctrica facturada total.
With the current capabilities of computers, Machine Learning algorithms are easily implemented in different areas of interest. In the present study, they are used to analyze the monthly billed electricity data in the city of Buenos Aires, during the period 2010-2021. The objectives are to determine patterns in the data using the K-Means algorithm and to determine the variables that most impact the total billed energy using the Linear Regression algorithm. Principal component analysis was used as a dimensionality reduction technique. The research was of a quantitative-explanatory type using data from the General Directorate of Statistics and Censuses of Buenos Aires, which were analyzed and preprocessed before applying the algorithms. To generate the models, 75% of the data is taken for training and 25% for the evaluation of the obtained model. For the K-Means grouping model, the optimal K was determined through the elbow method, and it was obtained that the total billed energy data present a monthly seasonality. For the Linear Regression model, the metrics R2, RMSE and MAE were used, and it was obtained that the residential, commercial, and industrial billed energies, plus the number of residential users, are the variables that have the greatest impact on the total billed electrical energy.
With the current capabilities of computers, Machine Learning algorithms are easily implemented in different areas of interest. In the present study, they are used to analyze the monthly billed electricity data in the city of Buenos Aires, during the period 2010-2021. The objectives are to determine patterns in the data using the K-Means algorithm and to determine the variables that most impact the total billed energy using the Linear Regression algorithm. Principal component analysis was used as a dimensionality reduction technique. The research was of a quantitative-explanatory type using data from the General Directorate of Statistics and Censuses of Buenos Aires, which were analyzed and preprocessed before applying the algorithms. To generate the models, 75% of the data is taken for training and 25% for the evaluation of the obtained model. For the K-Means grouping model, the optimal K was determined through the elbow method, and it was obtained that the total billed energy data present a monthly seasonality. For the Linear Regression model, the metrics R2, RMSE and MAE were used, and it was obtained that the residential, commercial, and industrial billed energies, plus the number of residential users, are the variables that have the greatest impact on the total billed electrical energy.
Description
Type
info:eu-repo/semantics/article
info:eu-repo/semantics/publishedVersion
Art.
info:eu-repo/semantics/publishedVersion
Art.
Source
Science, Engineering and Applications; Vol. 5 No. 2 (2022): Science, Engineering and Applications; 7-37
Ciencia, Ingenierías y Aplicaciones; Vol. 5 Núm. 2 (2022): Ciencia, Ingenierías y Aplicaciones; 7-37
2636-2171
2636-218X
10.22206/cyap.2022.v5i2
Ciencia, Ingenierías y Aplicaciones; Vol. 5 Núm. 2 (2022): Ciencia, Ingenierías y Aplicaciones; 7-37
2636-2171
2636-218X
10.22206/cyap.2022.v5i2