Author(s): TAN Xin, SHANG Yan-bo, HE Fa-yu, LIU Long-li, School of Resources & Civil Engineering, Northeastern University, State Key Laboratory of Mineral Processing, Beijing General Research Institute of Mining & Metallurgy, China Minmetals Corporation
Pages: 1198-1202
Year: 2016 Issue:  8
Journal: Journal of Northeastern University(Natural Science)
Keyword:  QSAR; hydroximic acid collector; genetic algorithm; selectivity; descriptor;
Abstract: Quantitative structure-activity relationships(QSAR) of 13 kinds of hydroximic acid collectors were studied to predict the flotation selectivity of cassiterite flotation separation based on genetic algorithm.Using fast descriptor and quantum chemical descriptor,two kinds of two-parameter models were developed,i.e.,Model 5 for selectivity index 1 and Model 6 for selectivity index 2.The correlative coefficients of Model 5 and Model 6 were 0.868 and 0.796,with R2 cvof 0.779 and 0.590,respectively.For the two models,internal reliability was validated using the leave-one-out approach,with R2= 0.785 for Model 5 and R2= 0.615 for Model 6,and an external validation was also performed using test sets and the mean absolute deviations were0.137 and 1.939,respectively.The results showthat Model 5 for selectivity index 1 has better predictability,and can effectively predict flotation selectivity of hydroximic acid collectors.