Summary

Spatial interpolation is a common and often used method in geoinformatics. There are many different interpolation methods, each of them used for different purposes. Artificial neural networks are popular computation method used in various scientific areas including geoinformatics.

The aim of this thesis was to test the use of module ann.* in GRASS GIS for spatial interpolation and to compare it with common interpolation techniques IDW and ordinary kriging. This module was also compared with neural networks packages nnet and neuralnet in R Project. The evaluation of methods was based mainly on RMSE, although time demands and user experience was also shown.

Multilayer perceptron with backpropagation training algorithm was used to perform the interpolation both in GRASS GIS and R Project. All the tests were done on artificial data created in R Project which simulated three surfaces with different characteristics. The final interpolation was made once for each of the surfaces. These data were split into two parts: training data (70 %) and testing data (30 %).

In order to find the best configuration for the multilayer perceptron many different settings were tested. Number of neurons in hidden layers was the main tested parameter. Then the interpolation was done. In GRASS GIS two slightly different approaches were used for interpolation using multilayer perceptron. The first model of neural network was trained on raster data file. This was the original use of ann.* module. Then a vector data file was prepared with the use of new script in Java (not a part of the ann.* module) and a new neural network model was trained on this file. After the multilayer perceptron models were trained interpolation was made in the same way with both networks.

The results indicate that multilayer perceptron in the ann.* module can be used for spatial interpolation purposes. However the resulting RMSE was higher then RMSE from IDW and ordinary kriging methods. Also the time demands were higher when using the neural networks ann.* module. When compared with neural network packages in R Project it is better to use the packages in R Project. Training of multilayer perceptron was faster in this case and results were the same or slightly better. However the resulting RMSE was still higher than the RMSE of the other methods (IDW, ordinary kriging). Also the use of neural networks is difficult for inexperienced users.

All the test were done on artificial data. If the real data were used, the results would be probably different. Also adding more input parametres could have improved the performance of the multilayer perceptron model.

According to the author of the ann.* module a new modules for vector data will be prepared (Netzel, 2011). The date of completion is unknown but it is possible that this new modules will be better for the purpose of spatial interpolation than the recent one. The ann.* module tested in this thesis cannot be considered as an equal replacement of the classical interpolation techniques, but it has a potential.

  • autor:
  • Veronika Nevtípilová
  • vedoucí práce:
  • Mgr. Justyna Pastwa
  • kontakt na autora:
  • v.nevtipilova@gmail.com
  •  
  • text práce v pdf

Tato práce byla zpracována v rámci bakalářského studia na Katedře geoinformatiky přírodovědecké fakulty Univerzity Palackého v roce 2013.