A hybrid approach to water potability prediction: leveraging artificial fish swarm algorithm and convolutional neural networks

Abstract

This study presents a novel hybridized analytic technique to solve the requirement of ensuring water quality within environmental engineering systems. The study addresses the significant issue of accurately determining water potability, which is crucial for public health, by combining the Artificial Fish Swarm Algorithm (AFSA) with Convolutional Neural Networks (CNNs). The collaboration between AFSA’s expertise in optimization and CNN’s capability in identifying patterns resulted in significant advancements in predicting accuracy. The independent CNN models demonstrated a notable accuracy of 95.73%. However, the suggested composite framework surpassed this performance by achieving a remarkable accuracy of 99.80%, resulting in a significant increase of 4.07% in precision. Furthermore, the precision and recall of the hybrid model reached a significant value of 99.73%. An examination of AFSA through analytical means has demonstrated that there exists a correlation between moderate step sizes and optimal algorithm performance. Furthermore, this analysis has revealed a notable behavioral adaptation from individual predation to collective swarming inside the AFSA system. The findings not only enhance the algorithm's resilience but also demonstrate its potential for proactive evaluation of water quality.