Estimating Bathing Water Quality from Meteorological Measurements

Traditional monitoring of bathing water quality based on the determination of faecal indicator bacteria is a time-consuming and costly process that always provides delayed information on water quality. In this study, an initial predictive model is created to estimate the quality of bathing water on the basis of meteorological data. The results show that for 2 out of 3 monitored bathing waters there is an imbalance between the number of data with acceptable and unacceptable water quality. However, overall results for water quality estimation reach accuracy of ∼ 60% with informedness ∼ 20% of cases. Comparing used algorithms, logistic regression has shown the highest combine metrics results on our dataset, compared to random forest and support vector machine.