An IoT-Based Active Learning Convolutional Neural Networks Model for Predicting River Water Quality

Abstract

The conventional procedures used to check river water quality involve slow processes that require significant expenses and fail to deliver immediate results. Without automated predictive systems, there are no capabilities to detect pollution trends ahead of time. The goal of this investigation is to establish an Internet of Things (IoT)-driven deep learning model that conducts real-time river water quality predictions through Deep Neural Network (DNN) algorithms with enhanced precision and effectiveness. This study uses an IoT system to acquire pH parameters and a DNN to predict water quality in the Tigris River of Iraq. Subsequently, an IoT-based Active Learning Convolutional Neural Network (IoT-ALCNN) model is developed to predict river water quality. The model is trained by a backpropagation algorithm that has an active learning mechanism. This mechanism provides an interface to enable the user to update the training data to improve the learning procedure. The performance of the IoT-ALCNN model is assessed through the use of classification assessment metrics. The performance of the model fit to the data is evaluated through the relationship between the residuals and updated values. The test results of the IoT-ALCNN model show that it can be used as a method for predicting and monitoring river water quality, as it achieves the best accuracy of 98.72% and outperforms other baseline models of both generalization and efficiency