Prediction of Alum Dosage in Coagulation Process by Weka Program: A Case Study in Vientiane Capital, Lao PDR

Abstract

This research topic entitled of "Prediction of Alum Dosage in the Coagulation Process using Weka Program". The data resources collected from 2 water treatment plants i.e. Chinaimo Water Treatment Plant (CWTP) and Dongmarkkaiy Water Treatment Plant in Vientiane Capital, Lao PDR. Those data resources were collected from the previous Jar-Test experimental. For the CWTP, the data resources collected from 2009 to 2016, we selected 2,038 records. For the DWTP, the data resources collected from 2008 to 2016, we selected 2,802 records. The model building for alum dosage prediction, we used 4 methods i.e. Multilayer Perceptron (MLP), M5Rules, M5P, and REPTree with 2 data groups i.e. the first data group, we substituted all missing values of each parameter by the average values of that parameter. For the second data group, we have cut off the missing values to reduce bias. The results indicated that the model building for alum dosage prediction by M5Rules method from the model group 1 of the CWTP realizes the less RMSE of 4.043 than another method when we have used this method to predict the alum dosage in the real applications. Thus, the M5Rules method realizes higher precision and credibility than other methods. On the other hand, in the DWTP, we found that the model building for alum dosage prediction by using MLP method from model group 1 realizes the less RMSE of 1.849 when we used it to predict the alum dosage in the real applications. Therefore, the MLP method yields the higher accuracy and dependability than other methods of the DWTP. Finally, the model had the highest precision in the drying season than raining season.