Comparison of Logistic Regression and Artificial Neural Network Algorithms in Land Cover Transition Potential Empirical Modeling of Coastal Areas of Mazindaran Province

Land cover changes and residential developments result in destruction of natural habitats and biodiversity. Land cover models are one of the most important methods to evaluate this trend. The objective of this study was comparison of Logistic Regression (LR) and Multi-Layer Perceptron Artificial Neural Networks (MLP ANNs) algorithms for transition potential modeling of coastal areas of Mazindaran province. Landsat satellite imageries, specifically 1988, 2001, 2006, and 2010 were used for change analysis. In addition, transition potential modeling was conducted using a logistic regression and multi layer perceptron artificial neural network. Each calibration period 2001–2006 was examined using Markov chain and hard prediction for extrapolating the year 2010. The accuracy assessment model was determined by kappa index. The results showed that logistic regression (0.8456) was more accurate than the multi layer perceptron artificial neural network (0.8276) in this study area.