Comparison of decision trees and artificial neural networks in estimating of wastewater and active sludge characteristics

In determining the properties of wastewater the amounts of biochemical oxygen demand (BOD5), chemical oxygen demand (COD), total organic carbon (TOC) and dissolved oxygen (DO) are the most basic measurement criteria for characterization of wastewater. Biological oxygen demand analysis (BOD5), together with the analysis of acidity (pH), temperature (T), conductivity (C), dissolved oxygen (DO), oxygen saturation (SO), salinity (SA), electrical conductivity (EC), chemical oxygen demand (COD), suspended solids (LSS), total nitrogen (TN) and total phosphorus (TP) made for the samples taken from the raw waste water coming to waste water treatment plants or treated waste water, lasts at least 5 days, as all others less than a day. In a study in which the above parameters were measured before, the effects of these parameters in the data set of 334 samples on the BOD5 parameter were investigated by using the decision tree method by the KNIME data mining package. Thus, taking into account the weighted effects of the parameters whose effects on the BOD5 parameter are known, the probable BOD5 value of an unknown sample has been estimated. In this study, based on this data set, Decision Trees and Artificial Neural Networks, which are among the data mining methods, were examined in detail in terms of both structural and results. When the results of both methods are compared, it could be seen that the distributions among the classes in binned values are close, but except for minor differences. It should be kept in mind that these shifts could be partially eliminated when the number of classes is increased. In addition, these results can be optimized in future studies by changing parameters such as the number of groupings or gain for (Decision Trees), and such as network layer number and gain rate for (Artificial Neural Networks).