Classification Model for Forest Fire Hotspot Occurrences Prediction Using ANFIS Algorithm

3rd International Symposium on LAPAN-IPB Satellite For Food Security and Environmental Monitoring 2016

地球科学;轻工业;生态环境科学

Wijayanto, A.K.^1 ; Sani, O.^2 ; Kartika, N.D.^2,3 ; Herdiyeni, Y.^4

Center for Environmental Research, Bogor Agricultural University, Kampus IPB Darmaga, Gedung PPLH Lt 3, Jl Lingkar Akademik, Bogor

16680, Indonesia^1

Master of Science in Information Technology for Natural Resources Management, Bogor Agricultural University, SEAMEO BIOTROP Campus, km 6 Tajur, West Java, Bogor

16721, Indonesia^2

The Agency of the Assessment and Application of Technology (BPPT), Jakarta, Indonesia^3

Computational Intelligence Lab, Department of Computer Science, Bogor Agricultural University, Darmaga Campus, Wing 20, West Java, Bogor

16680, Indonesia^4

关键词: Adaptive neuro fuzzy inference systems (ANFIS);    Classification models;    Early Warning System;    Fuzzy Inference systems (FIS);    Input-output data;    Nonlinear mappings;    Soft computing methods;    Spatial attribute;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/54/1/012059/pdf
DOI  :  10.1088/1755-1315/54/1/012059

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

This study proposed the application of data mining technique namely Adaptive Neuro-Fuzzy inference system (ANFIS) on forest fires hotspot data to develop classification models for hotspots occurrence in Central Kalimantan. Hotspot is a point that is indicated as the location of fires. In this study, hotspot distribution is categorized as true alarm and false alarm. ANFIS is a soft computing method in which a given inputoutput data set is expressed in a fuzzy inference system (FIS). The FIS implements a nonlinear mapping from its input space to the output space. The method of this study classified hotspots as target objects by correlating spatial attributes data using three folds in ANFIS algorithm to obtain the best model. The best result obtained from the 3rd fold provided low error for training (error = 0.0093676) and also low error testing result (error = 0.0093676). Attribute of distance to road is the most determining factor that influences the probability of true and false alarm where the level of human activities in this attribute is higher. This classification model can be used to develop early warning system of forest fire.