Induction of Model Trees for Predicting BOD in River Water: A Data Mining Perspective

摘要

Water is a primary natural resource and its quality is negatively affected by various anthropogenic activities. Deterioration of water bodies has triggered serious management efforts by many countries. BOD is an important water quality parameter as it measures the amount of biodegradable organic matter in water. Testing for BOD is a time-consuming task as it takes 5 days from data collection to analyzing with lengthy incubation of samples. Also, interpolations of BOD results and their implications are mired in uncertainties. So, there is a need for suitable secondary (indirect) method for predicting BOD. A model tree for predicting BOD in river water from a data mining perspective is proposed in this paper. The proposed model is also compared with two other tree based predictive methods namely decision stump and regression trees. The predictive accuracy of the models is evaluated using two metrics namely correlation coefficient and RMSE. Results show that the model tree has a correlation coefficient of 0.9397 which is higher than the other two methods. It also has the least RMSE of 0.5339 among these models.