A novel PCA-based approach for building on-board sensor classifiers for water contaminant detection

Water pollution causes an ever-increasing number of diseases and represents a worldwide concern, both for governments and researchers, as well as public opinion. This pollution also regards drinkable water, with two billion people plagued by this problem. Therefore, it is crucial to find reliable and low-cost technologies for a continuous and diffused monitoring of water. In this paper, we present a novel approach that allows the detection of water contaminants by using an ad-hoc classification system that can be implemented aboard low-cost sensors. To this aim, we first project the input data from the sensors into a 3-D space by using the PCA algorithm, then we use an ad-hoc devised classifier to distinguish the contaminants in the transformed space. We used an evolutionary algorithm to learn the parameters of the classifiers. The experiments were performed on a large dataset containing data from four contaminants, with the phosphoric and sulphuric acids, among the others. The results obtained confirm the effectiveness of the proposed approach.