NEURO-FUZZY MODELING: A PROMISING ALTERNATIVE FOR RISK ANALYSIS IN URBAN AFFORESTATION MANAGEMENT

Abstract

Urban afforestation has important functions, but problems related to its management are equally relevant, analysis of which is needed in order to prevent accidents. However, due to the subjectivity in the assessment, there may be uncertainty as to the seriousness of the risk. In order to address this, the present work evaluates a neuro-fuzzy-based methodology for the integrated analysis of risk indicators. From the knowledge of experts and a database with 107 cases, systems were constructed for the multi-criteria analysis of 18 parameters integrated using 3 indexes and 5 indicators. As a result, the model presented accuracies of 95.5% in generalization tests, and almost perfect agreement (kappa > 0.8) with the assessment by the expert. In conclusion, the findings show that this neuro-fuzzy modeling approach represents a promising alternative for supporting risk analysis in urban afforestation.