Predicting areas with ecotourism capability using artificial neural networks and linear discriminant analysis (case study: Arasbaran Protected Area, Iran)

In this study, the common systematic approach in Iran as well as a multilayer perceptron neural network were used to evaluate the ecological capability of the area for ecotourism. The performance of the artificial neural network (ANN) and linear discriminant analysis (LDA) method in the prediction and ranking of areas with ecotourism capability were also compared. Based on the results obtained, the ANN with an overall accuracy of 97% outperformed LDA (overall accuracy of 86%) in terms of the prediction and classification of recreational areas. Therefore, for each class, the ANN with an accuracy, precision, and sensitivity of 98%, 94.33%, and 86.67%, respectively, outperformed the LDA with the corresponding values of 90.67%, 55.33%, and 40.33%, respectively. Based on the ANN-modeled map, 0.17%, 10.09%, and 89.74% of the area were shown to belong to intensive recreation class 2, extensive recreation class 2, and the not suitable for recreation class, respectively. Therefore, the ANN functions well with higher accuracy for modeling and classification of areas with ecotourism capability compared to LDA.