Classification and Analysis of Water Quality Using Machine Learning Algorithms

Abstract

This research work revolves around the development of supervised machine learning models that can automatically classify the quality of river water. The original dataset is transformed and binned into two (swimming, boating) class types. Using this data an exploratory study of the machine learning models has been done as to construct a generic water quality classifier. At the same time, it was found that there is an imbalance in dataset. To overcome this problem SMOTE algorithm was applied and the exploratory analysis of the machine learning algorithms was done. The performance analysis of the various classifier algorithms show initially that there is a need for customization of the machine learning so that generalized classifier can be build. Deeper analysis of the study showed that the correlation-based features are helpful in RF and CART. At the same time, the PCA data projection shows a higher level of accuracy (0.989) with the neural network algorithm