Classifier for Drinking Water Quality in Real Time

ABSTRACT:

        The general quality of water in lakes, rivers and coastal areas is periodically assessed by scientific and environmental institutions. Real time features are critical for automatic assessment of Drinking Water Quality (DWQ). This paper explores the use of real time features to feed machine learning classifiers for DWQ. Two different representative datasets were used from: a) The Provincial Water Quality Monitoring Network from Ontario, Canada and b) National Hydrologic Information System from Central Region of Portugal. The procedure followed in this study was: (1) automatically computing a Water Quality Index to classify the datasets elements in five classes (excellent, good, medium, bad and very bad) using the Kumar method; (2) selecting best performed real time features on results of classified datasets; and (3) exploring machine learning algorithms (e.g. Decision Trees, Artificial Neural Networks and k-Nearest Neighbor) for producing DWQ classifiers. In this work, we perform the classification of two classes (good and medium) out of the five possible categories, due to the absence of vectors in the datasets.