An Efficient Feature Selection with Weighted Extreme Learning Machine for Water Quality Prediction and Classification Model

Abstract

Advanced urbanization and industrialization have resulted to a worsening of water quality significantly, and led to severe diseases. Water quality index (WQI) is a commonly employed measure to determine water quality, which is a costlier and lengthy process. The increasing effect of poor water quality leads to the requirement of latest machine learning (ML) models for automated water quality prediction. This paper aims to present a new feature selection with classification model for the proficient prediction of water quality in real time scenarios.  The presented model involves three processes, such as preprocessing, feature selection, and classification. Primarily, the dataset is collected and preprocessing takes place to transform the actual dataset into a compatible format for classification process. The proposed method uses a quantum teaching and learning based optimization (TLBO) algorithm to select an optimal set of features, and thereby reduces the complexity level. Besides, the presented QTLBO-WELM model also uses a weighted extreme learning machine (WELM) model for classification process. In order to assess the predictive outcome of the presented model, a series of experiments were conducted on a collection of 35 groundwater samples from Dharmapuri district in Tamil Nadu. The experimental values showcased the betterment of the presented QTLBO-WELM model with the sensitivity of 96.29%, specificity of 94.10%, accuracy of 95.71%, precision of 98.11%, F-score of 97.17%, and kappa value of 88.82%.