Chicken moth flame optimization and region-based convolution neural network for water quality prediction

Water is an important source for the sustenance of life, and its quality has a direct impact on the environment and public health. Water is utilized for various practices, such as agriculture, industry, and drinking. Geo-environmental pollution caused by various types of waste such as municipal, industrial, medical, solid, and agricultural fields makes the water unsuitable for usage. Water quality is primarily impacted by the discharge of agricultural and industrial effluents into the environment, which disrupts biological systems. Predicting water quality is crucial for environmental monitoring, ecosystem sustainability, and aquaculture. Accurate water quality prediction is essential for sustainable water management. Hence, the quality of the water should be maximized by managing water resources. In this research, an optimization-enabled deep learning model named chicken moth flame–region-based convolution neural network (CMF-RCNN) is introduced to predict water quality. Here, hidden properties of water are analyzed and utilized to predict the water quality characteristics. Moreover, input data are normalized using Z-score normalization and optimal features are selected via correlation analysis. Later, the water quality is predicted from the selected features using RCNN. The prediction performance of RCNN is enhanced by fine-tuning its weights by utilizing CMF. Moreover, the performance of CMF-RCNN is analyzed with respect to existing water quality prediction models, and the CMF-RCNN attained superior performance with precision of 0.927, recall of 0.946, and F1-score of 0.936, respectively.