Hybrid NCA-MRMR and Mamdani Type 1 Fuzzy Logic-Based Classification of Chemical Contamination State in Marine Transitional Systems

Water pollution is one of the pressing issues of the world that needs to be resolved. It must be properly monitored in order to address issues efficiently and to prevent it from negatively affecting the aquatic ecosystem and its surroundings. As such, this study proposes a Mamdani fuzzy-based system and ANFIS for chemical contamination level classification in the marine transitional systems of Portugal. The classification was done using four significant trace elements, namely arsenic, cadmium, chromium, and mercury concentrations in mg/kg. For the fuzzy logic model, each variable had three linguistic values (low, middle, high). It also had 81 rules in total. As for the ANFIS model, its initial FIS was generated using a dataset and the grid partition method. Hybrid learning algorithm with 30 epochs was then used to train and optimize this FIS. Both models were developed using MATLAB. Overall, the fuzzy logic model performed better than the ANFIS model, with an accuracy of 96.43% which was higher by 3.57%. Further improvements can be made by training the model using a bigger dataset. The proposed fuzzy system can also be integrated into monitoring systems to ensure that chemical status of aquatic systems is within acceptable ranges.