MLP network

This study presents an extension of the Gaussian process regression model for multiple-input multiple-output forecasting. This approach allows modelling the cross-dependencies between a given set of input variables and generating a vectorial prediction. Making use of the existing correlations in international tourism demand to all seventeen regions of Spain, the performance of the proposed model is assessed in a multiple-step-ahead forecasting comparison. The results of the experiment in a multivariate setting show that the Gaussian process regression model significantly improves the forecasting accuracy of a multi-layer perceptron neural network used as a benchmark. The results reveal that incorporating the connections between different markets in the modelling process may prove very useful to refine predictions at a regional level.

Rangelands production measurement is time-consuming and expensive. Therefore, models are often employed to simulate rangelands conditions as a supplement. Artificial neural network (ANN) is widely used for modeling in environmental studies, yet it cannot preset its results in the form of a map or geo-referenced data. We used ANN to estimate the spatial distribution of rangelands production, then a geographic information system (GIS) was applied as a pre-processing and post-processing framework in rangelands production modeling. The ANN was trained (Rsqr = 0.95, MSE = 0.02) and tested using data from the Baladeh rangelands located in the northern part of Iran. Rangelands production was simulated using a multi-layer perceptron (MLP) network. We estimated rangelands production (using many plots and field studies) as the network output, along with the influencing factors in the production (vegetation, climatic, topographic, edaphic and human factors) as the inputs. After modeling and model optimizing in ANN, the model test was performed (Rsqr=0.8, MSE=0.3). Furthermore, the studied area was divided with the pixels 100×100 m (raster format) in the GIS medium. Then, the digital layers of the network inputs were combined and a raster layer was prepared including the network inputs values and geographic coordinate. The values of pixels (network inputs) were imported in ANN (NeuroSolutions software). Rangelands production was simulated using the validated optimum network in the sites without production measurements. In the next step, the results of ANN simulation were imported in the GIS medium, then rangelands production map was prepared based on the estimated results of ANN. The results indicated that integrating ANN and GIS exhibits high accuracy and performance in rangelands production estimation. Hence, the prepared rangelands production map can be used for planning and managing the rangelands.