Open AccessPredicting Fish Ecological As Indicator of River Pollution Using Decision Tree Technique
Author(s) -
Che-Yu Hsu,
Sin-Liang Ou,
Wei-Fan Hsieh
Publication year - 2018
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.179
H-Index - 26
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/164/1/012022
Subject(s) - decision tree , water quality , chaid , logistic regression , pollution , index of biological integrity , index (typography) , environmental science , ecology , diversity index , species richness , hydrology (agriculture) , computer science , data mining , machine learning , engineering , geotechnical engineering , world wide web , biology
The Goal of the research is to introducing the principle of Decision Tree that is being used to forecast river pollution, it provides a new method to evaluate the river pollution based on its water quality. We collected monthly monitoring data of water quality from Dezikou River basin of Yilan County, and the data of fish ecology obtained from ecological survey and report, in which to build a water quality and ecology resources database through an actual field investigation. By using data mining software, IBM SPSS Modeler 14.1’s decision tree, conducting the River Pollution Index. Shannon-Weaver diversity, Pielou’s Evenness Index, Margalef’s Species Richness Index, Fish Tolerant Index and Simpson’s Index of Diversity’s classification and prediction, to build a model for river pollution prediction, and to compare this with the Multiple Logistic Regression Analysis. The results showed that the model for river pollution prediction built under the Decision Tree can obtain a better forecast result. The following are the accurate rates of Decision Tree: 88% for CART, 90% for CHAID, 91.67% for C5.0, and 86.11% for Multiple Logistic Regression Analysis. Therefore, the Decision Tree’s algorithm shows a better result in forecasting than the Multiple Logistic Regression Analysis.