Comparison of Procedure of Artificial Neural Networks, Logistic Regression and Similarity Weighted Instance-Based Learning in Modeling and Predicting the Destruction of the Forest (Case Study: Gorgan-Rood Watershed- Golestan Province)

The change in forest cover plays a vital role in ecosystem services, atmospheric carbon balance and thus climate change. The goal of this study is comparison of three procedure of Artificial Neural Network, Logistic regression and Similarity weighted Instance-based Learning (SIM Weight) to predict spatial trend of forest cover change.

In this study, land use maps for the periods 1984 and 2012 derived from Landsat TM satellite imagery, was used. Transition potential modeling using artificial neural network, Logistic regression and Similarity weighted Instance-based Learning and prediction based on the best model using Markov chain model was performed. In order to assess the accuracy of modeling, statistics of relative performance characteristic (ROC), ratio Hits/False Alarms and figure of merit was used.

The results show the accuracy of artificial neural network with the ROC equal to 0.975, the ratio Hits/False Alarms equal to 63 percent and the figure of merit is equal to 12 percent.

 Discussion and Conclusions

Artificial Neural Networks in comparison with Logistic Regression and Similarity weighted Instance-based Learning has higher accuracy and less error in modeling and predicting of forest changes.