Estimating Water Quality using Internet of Things and Machine Learning

Abstract

Quality of water plays an important role in all aspects of our lives and it has been deteriorating at an alarming rate due to pollution, deeming its quick, inexpensive and accurate detection vital. Conventional methods to calculate water quality are lengthy, expensive and inefficient. This thesis reviews the conventional lab analysis methods of determining water quality to gain insight into the problem, state of the art machine learning methodologies and role of IoT in determining water quality more efficiently. Also, this thesis proposes a method to detect and predict water quality in real time, respectively, using IoT and machine learning. This thesis explores several machine learning algorithms and predicts water quality using minimal and easily attainable water quality parameters i.e. Temperature, pH, Turbidity and Total dissolved solids. Logistic Regression algorithm yields the most accurate results with accuracy up to 77.98% without TDS and accuracy up to 84.01% with TDS.