Predicting Size of Forest Fire Using Hybrid Model

Shidik, Guruh Fajar; Mustofa, Khabib

doi:10.1007/978-3-642-55032-4_31

Cited by 22 publications

(12 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zwirglmaier et al (2013) used a BN to predict area-burned classes using historical fire data, fire weather data, fire behavior indices, land cover, and topographic data. Shidik and Mustofa (2014) used a hybrid model (fuzzy C-means and back-propagation ANN) to estimate fire size classes using data from Cortez and Morais (2007); the hybrid model performed best, with an accuracy of 97.50% when compared with naïve Bayes (55.5%), DT (86.5%), RF (73.1%), KNN (85.5%), and SVM (90.3%). Mitsopoulos and Mallinis (2017) compared BRT, RF, and LR to predict three burned-area classes for fires in Greece.…”

Section: Burned-area and Fire-severity Predictionmentioning

confidence: 99%

A review of machine learning applications in wildfire science and management

et al. 2020

View full text Add to dashboard Cite

Artiﬁcial intelligence has been applied in wildﬁre science and management since the 1990s, with early applications including neural networks and expert systems. Since then the ﬁeld has rapidly progressed congruently with the wide adoption of machine learning (ML) methods in the environmental sciences. Here, we present a scoping review of ML applications in wildﬁre science and management. Our overall objective is to improve awareness of ML methods among wildﬁre researchers and managers, as well as illustrate the diverse and challenging range of problems in wildﬁre science available to ML data scientists. To that end, we ﬁrst present an overview of popular ML approaches used in wildﬁre science to date, and then review the use of ML in wildﬁre science as broadly categorized into six problem domains, including: 1) fuels characterization, ﬁre detection, and mapping; 2) ﬁre weather and climate change; 3) ﬁre occurrence, susceptibility, and risk; 4) ﬁre behavior prediction; 5) ﬁre eﬀects; and 6) ﬁre management. Furthermore, we discuss the advantages and limitations of various ML approaches relating to data size, computational requirements, generalizability, and interpretability, as well as identify opportunities for future advances in the science and management of wildﬁres within a data science context. In total, we identfied 300 relevant publications up to the end of 2019, where the most frequently used ML methods across problem domains included random forests, MaxEnt, artiﬁcial neural networks, decision trees, support vector machines, and genetic algorithms. As such, there exists opportunities to apply more current ML methods — including deep learning and agent based learning — in the wildﬁre sciences, especially in instances involving very large multivariate datasets. We must recognize, however, that despite the ability of ML methods to learn on their own, expertise in wildﬁre science is necessary to ensure realistic modelling of ﬁre processes across multiple scales, while the complexity of some ML methods, such as deep learning, requires a dedicated and sophisticated knowledge of their application. Finally, we stress that the wildﬁre research and management communities play an active role in providing relevant, high quality, and freely available wildﬁre data for use by practitioners of ML methods.

show abstract

Section: Burned-area and Fire-severity Predictionmentioning

confidence: 99%

A review of machine learning applications in wildfire science and management

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Several computational intelligence approaches were used namely: Multilayer Perceptron (MLP), Radial Basis Function Networks (RBFN), Support Vector Machines (SVM) and fuzzy logic. Shidik and Mustofa [18] used a Back-Propagation Neural Network which was trained based on meteorological and forest weather indices, so as to classify the burned area in three categories. Aldrich et al [1] investigated the effect of variations in land use and climate in the occurrence of forest fires.…”

Section: Literature Reviewmentioning

confidence: 99%

A Hybrid Soft Computing Approach Producing Robust Forest Fire Risk Indices

Anezakis

Demertzis

Iliadis

et al. 2016

IFIP Advances in Information and Communication Technology

View full text Add to dashboard Cite

Forest fires are one of the major natural disaster problems of the Mediterranean countries. Their prevention-effective fighting and especially the local prediction of the forest fire risk, requires the rational determination of the related factors and the development of a flexible system incorporating an intelligent inference mechanism. This is an enduring goal of the scientific community. This paper proposes an Intelligent Soft Computing Multivariable Analysis system (ISOCOMA) to determine effective wild fire risk indices. More specifically it involves a Takagi-Sugeno-Kang rule based fuzzy inference approach, that produces partial risk indices (PRI) per factor and per subject category. These PRI are unified by employing fuzzy conjunction T-Norms in order to develop pairs of risk indices (PARI). Through Chi Squared hypothesis testing, plus classification of the PARI and forest fire burned areas (in three classes) it was determined which PARI are closely related to the actual burned areas. Actually we have managed to determine which pairs of risk indices are able to determine the actual burned area for each case under study. Wild fire data related to specific features of each area in Greece were considered. The Soft computing approach proposed herein, was applied for the cases of Chania, and Ilia areas in Southern Greece and for Kefalonia island in the Ionian Sea, for the temporal period 1984-2004.

show abstract

“…Since the 1990s, the surge in the availability of data and covariates has spurred the use of these techniques to predict wildfire behaviour. Jain et al (2020) found 127 journal papers or conference proceedings published up to the end of 2019 on ML applied to fire occurrence, susceptibility and risk; of these adversarial neural networks (ANN) were the most prominent (e.g., Liang et al, 2019;Dutta et al, 2013;Shidik and Mustofa, 2014). For wildfire occurrences, most studies focus on classification tasks instead of count modelling.…”

Section: Introductionmentioning

confidence: 99%

“…Dutta et al (2013) explored different combinations of training-testing splits to identify the best possible paradigm to maximize the generalization capability of their ANN architecture. K-fold cross-validation is also popular (De Angelis et al, 2015;Shidik and Mustofa, 2014;Xie and Peng, 2019;Mitsopoulos and Mallinis, 2017), but it may give overly optimistic evaluations for spatially dependent data (Roberts et al, 2017). An alternative is spatial cross-validation (Pohjankukka et al, 2017), but it is still unclear how best to construct spatial folds in this context, and doing so anyway ignores time dependencies.…”

Section: Introductionmentioning

confidence: 99%

Gradient boosting with extreme-value theory for wildfire prediction

Koh¹

2021

Preprint

View full text Add to dashboard Cite

This paper details the approach of the team Kohrrelation in the 2021 Extreme Value Analysis data challenge, dealing with the prediction of wildfire counts and sizes over the contiguous US. Our approach uses ideas from extreme-value theory in a machine learning context with theoretically justified loss functions for gradient boosting. We devise a spatial cross-validation scheme and show that in our setting it provides a better proxy for test set performance than naive cross-validation. The predictions are benchmarked against boosting approaches with different loss functions, and perform competitively in terms of the score criterion, finally placing second in the competition ranking.

show abstract

Predicting Size of Forest Fire Using Hybrid Model

Cited by 22 publications

References 13 publications

A review of machine learning applications in wildfire science and management

A review of machine learning applications in wildfire science and management

A Hybrid Soft Computing Approach Producing Robust Forest Fire Risk Indices

Gradient boosting with extreme-value theory for wildfire prediction

Contact Info

Product

Resources

About