Water quality prediction based on Naïve Bayes algorithm.

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Increasing global water demand and pollution necessitate advanced water quality assessment methods.
• Traditional water quality prediction models struggle with large datasets and missing information.
• Machine learning (ML) offers a promising alternative for real-time water quality analysis.

Purpose of the Study:

• To apply the Naïve Bayes machine learning algorithm for predicting water quality classes.
• To develop and validate a predictive model using key water quality parameters.
• To assess the efficacy of ML in handling missing data for water quality management.

Main Methods:

• Utilized the Naïve Bayes algorithm, a common machine learning technique.
• Developed a prediction model incorporating nine water quality parameters: temperature, pH, electrical conductivity, oxygen saturation, biological oxygen demand, suspended solids, nitrogen oxides, orthophosphates, and ammonium.
• Trained and validated the model using 48 water samples from 2013-2019 across five locations in Vojvodina Province, Serbia, using Netica software.

Main Results:

• The Naïve Bayes model achieved a high prediction accuracy of 64 out of 68 cases.
• The model demonstrated effectiveness in predicting water quality classes even when presented with missing data points.
• The developed model proved to be a trustworthy tool for water quality assessment.

Conclusions:

• The Naïve Bayes machine learning model is a reliable and effective tool for real-time water quality prediction.
• This ML approach is particularly valuable for managing large datasets and addressing missing data challenges.
• The study supports the transition towards digital water management systems through advanced predictive analytics.