Water Quality Prediction and Classification using AFSO based Long Short-Term Model with Data Transformation Manuscript

Abstract

Water is a precious, essential, and dwindling resource in both developing and developed nations. As a vital nutrition for human beings, it easily takes the cake as the planet's most valuable natural resource. Various wastes, including municipal, industrial, agricultural (including pesticides and fertilizers), medical, etc., contribute to geo-environmental contamination and render water unfit for human or animal use. Therefore, it is crucial to establish effective means to automate water suitability checking. In this investigation, the variables included are pH, hardness, solids, chloramines, sulphate, conductivity, organic carbon, trihalomethanes, and turbidity. These measurements serve as a feature vector to represent the state of the water. The article employed an enhanced deep learning model (IDL) to predict the water quality class. Normalization, spitting, and transformation are three of the models used in preliminary data processing. Long Short-Term Memory (LSTM) models take in preprocessed data as input, and their weights are ideally chosen using Artificial Fish Swarm Optimization (AFSO). Using a dataset obtained from Kaggle, tests were conducted with several levels of granularity. The findings showed that the LSTM classifier is superior to the rest. The results show that deep learning methods may effectively forecast the viability of water quality.