Water Microbial Pollution Risk Simulation and Prediction using ML

Water quality has been a global concern due to its scarcity as well as it is direct effects on wellbeing and food production. Therefore, many studies have been addressing this matter in order to address the consequences left behind. Consider the harmful effects of polluted water moving uninterrupted over vast distances and through countless homes. The water we drink, cook with, and bathe in would not be safe requiring an immediate response in order to reduce the damage as a direct result of the pollution. The aim of this study was to develop models that could evaluate the microbial water contamination by understanding its pattern and thus predicting its behavior in advance based on spatiotemporal analysis and time series analysis. The goal of this study is to (a) understand the relevant factors that could precede the microbial contamination and their impact on spreading the contamination. (b) predict and forecast the future values of Coliform. Coliform and E.coli are types of bacterium where their presence in water is considered as a potential and/or extreme health hazard that may cause food poisoning. I have used and analyzed water quality parameters trends from 2015 to 2021 by using seasonal trend decomposition of time series plots in order to understand the behavior of Coliform through water distribution system allocating at different places across New York that are previously detected with E.coli. Three models are developed in order to predict the contamination with Coliform for all sites for the next 15 months. The machine learning algorithms are Holt-winters method, ETS, and seasonal ARIMA. Seasonal ARIMA model achieved better accuracy compared to other models.