OIL SPILL EMERGENCY DSS BASED ON ARTIFICIAL NEURAL NETWORKS

Oil spill cases are important environmental emergency accidents. The most important elements for response and treatment of an oil spill are "quickness" and "accuracy". This paper presents a methodology for developing oil spill emergency Decision Support System (DSS) based on Artificial Neural Network (ANN), in which cases for the training process of ANN are from a Case-Based Reasoning (CBR) system which was developed in our previous studies. Information of cases is divided into the input-attribute information and decision-making information, which is digitized (coded) accordingly. In the construction of oil spill ANN, a modified back-propagation algorithm is employed to do training. Output data of ANN are interpreted back to practical contingency measures through translation (decoding) of decision-making information. The training features, time, errors, accuracy, and input-attribute weights of the developed ANN system are analyzed. The research results show that: using cases from the CBR system for training, the developed ANN oil spill decision support system is effective, which can overcome shortcomings of ANN and CBR each other, and generate emergency plans quickly and accurately to meet the requirements of an oil spill on-site.