ISOLATION FOREST WITH OPTIMAL ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM BASED WATER QUALITY PREDICTION AND CLASSIFICATION MODEL

In recent times, the water shortage problem in semi-arid regions has reached to a critical stage. Water quality level estimation is found to be important for societal and economic development. The water quality index (WQI) is commonly employed to determine the threat to water quality and helps to attain effective water resource management. Since the traditional WQI calculations are tedious, time consuming and inconsistent, the recently developed machine learning (ML) techniques can be applied for water quality prediction. This paper presents a new ML based water quality prediction technique isolation forest (IF) with optimal adaptive neuro-fuzzy inference system (OANFIS) model. The presented IF-OANFIS model involves pre-processing at the earlier for converting the data into a compatible format. In addition, IF based outlier detection technique is employed to remove the outliers exist in the data. For prediction process, OANFIS classifier is used where the parameters of the ANFIS model are tuned by continuous particle swarm optimization (CPSO) algorithm. For experimental validation, the performance of the IF-OANFIS model is tested using 35 groundwater samples are collected from Dharmapuri district in Tamil Nadu and which were analyzed using different physical and chemical parameters. The simulation results verified that the IF-OANFIS model has attained maximum prediction performance with the sensitivity of 94.36%, specificity of 69.64%, accuracy of 88.76%, precision of 91.33%, F-score of 92.81%, and kappa value of 66.66%.