Dynamic Tuning of a Forest Fire Prediction Parallel Method
Fecha
2020-01-01Autor
Caymes Scutari, Paola
Tardivo, María
Bianchini, Germán
Méndez Garabetti, Miguel
Metadatos
Mostrar el registro completo del ítemResumen
Different parameters feed mathematical and/or empirical
models. However, the uncertainty (or lack of precision) present in such
parameters usually impacts in the quality of the output/recommendation
of prediction models. Fortunately, there exist uncertainty reduction
methods which enable the obtention of more accurate solutions. One
of such methods is ESSIM-DE (Evolutionary Statistical System with
Island Model and Differential Evolution), a general purpose method for
prediction and uncertainty reduction. ESSIM-DE has been used for the
forest fireline prediction, and it is based on statistical analysis, parallel
computing, and differential evolution. In this work, we enrich ESSIM-DE
with an automatic and dynamic tuning strategy, to adapt the generational
parameter of the evolutionary process in order to avoid premature
convergence and/or stagnation, and to improve the general performance
of the predictive tool. We describe the metrics, the tuning points and
actions, and we show the results for different controlled fires.
Colecciones
El ítem tiene asociados los siguientes ficheros de licencia: