Informace o publikaci
Identification of Pollution Sources by Machine-Learning Approach
| Autoři | |
|---|---|
| Rok publikování | 2007 |
| Druh | Článek ve sborníku |
| Konference | Proceedings of the International Symposium on Environmental Software Systems ISESS-2007 |
| Fakulta / Pracoviště MU | |
| Citace | |
| Klíčová slova | machine learning, clustering, classification, pollution, environmental data |
| Popis | A problem connected with measuring of chemical sets (pollution) at a certain Czech locality is investigated. The main goal is the automatic identification of pollution sources. Some sources produce the same mixtures of chemicals in different rates. To automatically determine what is the potential source of a specific combination of recorded pollutant mixtures would be very helpful especially for huge volumes of continuously recorded data. The used real-data collection comes from the period 1997-2005 of the high-volume sampling of ambience, specifically from the gas phase of pollutants: 468 measured samples, each described by 14 harmful attributes (chemicals). The combination of unsupervised clustering by the X-means algorithm followed by training a supervised classifier (e.g., RBF-networks, c4.5-trees) showed to be a very promising approach to sort out the problem with determination of typical pollutant sources. |