Comparison of AI-based Models Artificial Neural Network, Genetic Algorithm and Fuzzy Logic in Classifying Wetlands Regions in Indonesia

Abstract

food crop monitoring process. Advantages of using remote sensing technology in making estimates via object-based satellite images have significantly influenced decision-makers in monitoring crop producivity, subsequently helping them to formulate necessary steps in ensuring national food security.
The reliability of estimates as provided by satellite images data is highly dependent on the utilization of certain methods or models that can properly process the image’s data, so as to result in a high degree of accuracy of the estimation. In recent years, image extraction models that are based on Artificial Intelligence (AI) have been used to perform such estimations in favor of models based on statistical approach. Previous research has also noted that each extraction model has its own respective strengths and weaknesses.
This study attempts to compare the performance of different AI extraction models, namely Artificial Neural Network, Genetic Algorithm and Fuzzy Logic, in classifying regions of wetlands (rice fields) and non-wetlands in Indonesia. A comparison was also made against the statistical model of Maximum Likelihood in order to verify that AI-based models are generally more desirable than the statistical ones. A comparison based on mapping accuracy showed that the Genetic Algorithm model can be regarded as the best extraction model due to its highest value of mapping accuracy.