Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ Recolector de Cienci...arrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
addClaim

Modelos de predicción de incendios forestales

Authors: Baeza Ruiz-Henestrosa, Juan;

Modelos de predicción de incendios forestales

Abstract

En el presente trabajo fin de estudios se aborda el problema de la predicción diaria de incendios forestales en la Comunidad Autónoma de Andalucía haciendo uso de técnicas de procesamiento de datos espaciales y modelos de Machine Learning. Se fija el marco temporal del estudio entre los años 2002 y 2022. Se consideran 27 variables correspondientes a 6 categorías: antropogénica, meteorológica, topográfica, demográfica, hidrológica y de vegetación. Se usan los perímetros de incendios forestales mayores de 100 ha ocurridos en Andalucía y obtenidos a partir de imágenes satélite y datos de campo disponibles en la Red de Información Ambiental de Andalucía (REDIAM). Se implementan métodos para procesar los conjuntos de datos espaciales recopilados y generar muestras adecuadas para entrenar modelos predictivos, con los cuales se genera una muestra de 21.546 registros que se usa para entrenar los modelos, considerando una partición temporal en entrenamiento-validación-test. Los modelos analizados han sido: Regresión Logística con penalización, Regresión Logística con penalización usando PCA, k-Nearest Neighbours, SVM lineal, SVM radial, Árboles de Decisión y Random Forest. Se han ajustado los valores de los hiperparámetros evaluando el rendimiento sobre el conjunto de validación y se ha comparado el rendimiento de los modelos construidos sobre el conjunto test considerando diversas métricas. Han destacado los modelos de Regresión Logística lasso y SVM, que han obtenido los mejores resultados en el conjunto test. Finalmente, se ha evaluado el desempeño de estos modelos en dos casos prácticos, obteniendo resultados prometedores.

This undergraduate thesis addresses the problem of daily wildfire prediction in the Autonomous Community of Andalusia using spatial data processing techniques and Machine Learning models. The time frame of the study is set between 2002 and 2022. Twenty-seven variables are considered in the study belonging to six major categories: anthropogenic, meteorological, topographical, demographic, hydrological and vegetation. The perimeters of forest fires larger than 100 ha occurring in Andalusia and obtained from satellite images and field data available in the Environmental Information Network of Andalusia (REDIAM) are used. Methods are implemented to process the spatial datasets collected and generate adequate samples to train predictive models. Thus, a sample of 21,546 records is generated and used to train the models, considering a temporal partition in training-validation-test. The models analysed were: Logistic Regression with penalty, Logistic Regression with penalty using PCA, k-Nearest Neighbours, linear SVM, radial SVM, Decision Trees and Random Forest. Hyperparameter tuning has been carried out usig the validation set and the performance of the tuned models on the test set has been compared usig different metrics. The lasso Logistic Regression and SVM models stood out, achieving the best results in the test set. Finally, the performance of these models was evaluated in two case studies, yielding promising results.

Universidad de Sevilla. Doble Grado en Matemáticas y Estadística

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Green