Browsing by Author "DIB, Nadjet"

Now showing 1 - 1 of 1

Water quality classification using machine learning
(2024-10-06) DIB, Nadjet
A wastewater treatment plant (WWTP) is an essential part of the entire water cycle, which reduces concentrations of pollutants in the environment. To enhance the monitoring and control of WWTP efficiency , researchers developed different models and systems, This study presents the application of Machine learning-based (ML) Artificial intelligence techniques (AIT) such as Random Forest algorithm (RF), Support Vector Machine (SVM ) and Extreme Gradient Boosting (XGBoost) to design an automatic classifier for water quality and determine the appropriate destination for the treated wastewater ,providing justifications and direct recommendations based on international standards and thresholds, for this purpose, dataset consisting of 3600 values related to domestic WW was utilized, with the outputs categorized into two classes influent not pure water (untreated WW) and effluent pure water (treated WW) . Approximately 240 data points were sourced from Algerian records, spanning ten years of monthly data. The influent parameters including Biological Oxygen Demand (BOD), Chemical Oxygen Demand (COD), and Total Kjeldahl Nitrogen (TKN) were used as inputs for wastewater quality prediction, identified as the most predictive features through the correlationbased feature selection (CFS) method, sufficient data and correct values .The machine learning models were trained via 60% values of the dataset ,with their accuracy tested on the remaining 40%.From the results of the test .Random forest with the accuracy of 99,8% is found to be the most successful model although all models have excellent accuracy because in this case the effective features were just three and the data was simple, it seen that SVM model is the fastest technique although Random forest close results to SVM but it seems that the training speed of XGBoost is approximately 7 times longer than SVM. Moreover, different functions are then integrated to determine whether this predicted wastewater suitable for agriculture or environment or unsuitable for them both ,providing reasons and recommendations or advices which empower us to create a platform of digital water prediction by the implementation of machine learning coding ,The promising results obtained paved the way for forecasting the performance of WWTP operations by the prediction of water quality , optimizing the reuse of treated WW on agriculture and swiftly address process anomalies before they escalate into more severe issues thereby enabling informed decision-making by water system managers.