Estimation of Phosphorus Reduction from Wastewater by Artificial Neural Network, Random Forest and M5P Model Tree Approaches

Abstract

This study aims to examine the ability of free floating aquatic plants to remove phosphorus and to predict the reduction of phosphorus from rice mill wastewater using soft computing techniques. A mesocosm study was conducted at the mill premises under normal conditions, and reliable results were obtained. Four aquatic plants, namely water hyacinth, water lettuce, salvinia, and duckweed were used for this study. The growth of all the plants was inhibited in rice mill wastewater due to low pH, high chemical oxygen demand, high conductivity, and high phosphorus concentration. Subsequently, a 1:1 ratio of mill water to tap water was used. A control was maintained to assess the aquatic plant technology. In this study, the aquatic plants reduced the total phosphorus content up to 80 % within 15 days. A comparison between three modeling techniques e.g. Artificial neural network (ANN), Random forest (RF) and M5P has been done considering the reduction rate of total phosphorus as predicted variable. In this paper, the data set has been divided in two parts, 70 % is used to train the model and residual 30 % is used for testing of the model. Artificial neural network shows promising results as compared to random forest and M5P tree modelling. The root mean square error (RMSE) for all the three models is observed as 0.0162, 0.0204 and 0.0492 for ANN, RF and M5P tree, respectively.