Assessment of Water Quality in Smart City Environment Leveraging ML-IoT

Abstract

AbstractWater quality measurement and potability for human residents are critical for health concerns in smart cities. Citizens are perplexed by data about the quality of their drinking water derived from chemical measurements. As a result, data on water potability derived from chemical sensor values must be interpreted prior to being made publicly available. This paper proposed an edge-cloud ubiquitous sensor network for low-cost water quality measurement to supplement existing IoT-based infrastructure. Machine learning algorithms are applied to a dataset containing eight fields related to water potability. Following that, a total of 16 machine learning algorithms for potability prediction were compiled, including 11 shallow learning algorithms and 5 deep learning algorithms. The performance of multiple machine learning algorithms for determining the potability of water based on chemical and laboratory measurements was compared. These results were then compared to those obtained using deep learning algorithms such as ANN, CNN-Resnet, and CNN-LSTM. CNN-Batch Normalization, the most accurate of these algorithms, achieved a maximum testing accuracy of 85.03%.KeywordsPotable waterWater qualityMachine learningIoT