فهرست نشریات علمی دسترسی آزاد ایران

This study aimed to predict the spatial distribution of the tomato fruitworm, Helicoverpa armigera (Lep., Noctuidae), using an artificial neural network optimized with ant and artificial honeybee colony algorithms. Data on the population density of this pest were collected in a 2000 m2 tomato field located at the following geographical coordinates: 38S, 693942E, and 3800263N. In these models, latitude and longitude variables were used as input variables, and population changes in the tomato fruitworm larvae of different ages were used as output variables. The network used was a multilayer perceptron optimized using two metaheuristic algorithms. To evaluate the accuracy of the neural networks used to predict the spatial distribution of this pest, an average comparison was made between the spatially predicted values by the optimized neural network and their actual values. A comparison of the means showed no significant difference between the actual and predicted spatial datasets in the training and testing phases. A coefficient of determination of 0.9987 indicated that the neural network optimized with the artificial honey bee colony algorithm achieved a higher accuracy than the ant colony algorithm, with a coefficient of determination of 0.9911 for predicting the density of H. armigera moths. In addition, neural network-generated maps, optimized using both metaheuristic algorithms, showed that the pest's spatial distribution was cumulative