K-nearest neighbor based computational intelligence and RSM predictive models for extraction of Cadmium from contaminated soil

Abstract

• CI predictive models’ application on soil washing are scare in literature. • Predictive models faced dwindling accuracy when few sample data are available. • Performance of Cd ions soil extraction prediction using KNN were investigated. • KNN performance was better over traditional RSM for Cd soil extraction. • KNN are suitable for soil washing predictions under RSM few samples data. Computational intelligence (CI) predictive models based on k -Nearest Neighbor (KNN) algorithms were developed for Cd ions removal from contaminated soil using environmentally friendly chelating-agent polyaspartate. Based on extracted Cd ions into the chelating-agent, residual Cd ions in treated soil and Cd removal efficiency, the performances of the KNN models were compared with response surface methodology (RSM) models using whole data set (KNN1) and split data (KNN2) scenarios using correlation coefficient (R 2 ) and root mean square error (RMSE). Optimal performances of the developed KNN based models were found to be significantly influenced by the nearest neighbor’s k -parameter attributed to the disparity in the two approaches. The KNN1 demonstrated better performances characterized by higher R 2 = 0.984–0.999 and lower RSME of 0.399–6 against the RSM models’ R 2 = 0.7882–0.990 and RSME 2.08–20.36, respectively. For the KNN2 models, even though lower performances were obtained, yet the soil remediation efficiency models, demonstrated enhanced performance over the RSM models.