Abstract
This study proposes the use of Artificial Neural Networks (ANNs) to classify the feasibility of land leasing by sugar-energy plants in the Alta Paulista region. The model receives seven input variables—such as pH, organic matter, distance to the plant, lease cost, expected productivity, production cost, and the share of leasing in the total cost—to classify each sample as either “viable” (class 1) or “not viable” (class 2). The network architecture consists of one hidden layer with 10 neurons and two output neurons, responsible for the binary decision-making process. The data were collected through questionnaires answered by experts in the sector, forming the dataset used for training, validation, and testing of the ANN. The classification results are illustrated through confusion matrices. In the training set, the model achieved 100 % accuracy for both classes. During validation, an accuracy rate of 85 % was observed, with a few classification errors. The overall matrix (including all data) demonstrated excellent performance, with a total accuracy of 97 %, correctly classifying the majority of the samples. The model integrates agronomic and economic–operational variables from professionals and replicates contracting criteria, offering novelty not only in the ANN method but in its applied domain and decision-support value.