Environmental decision support system for Plane trees failure prediction: A comparison of multi‐layer perceptron and random forest modeling approaches

Abstract

AbstractTree risk is the likelihood of property damage or personal injury from a potentially hazardous tree in cities. We aimed to compare multilayer perceptron (MLP) and random forest (RF) model techniques to predict Plane trees failure hazard in urban areas. In this research, 500 Plane trees in different urban structures were selected, and 12 variables were measured for target trees prediction as model independent variables. Tree failure hazard (tree fails along next 3 yr) was modeled by two machine learning techniques: MLP and RF. The RF model represents the higher accuracy in training (100%), test (94.4%), and all data sets (98.4%). The results of variable importance calculation of designed RF model revealed that wind‐protected aspects, soil depth, length of cracks and cavities, and internal decay are the most important variables respectively that influence the classes of Plane tree failure. The urban foresters or designers can easily determine the possibility of Plane trees risk by running the designed graphical user interface and entering 12 variables of Plane tree. As a result, it will help foresters and urban area managers to make better decisions and predict the proper time to decrease harm from Plane tree failure.