Prediction of gold in Takab Zarshuran using linear and nonlinear regression methods

This paper show the importance of trace elements in the estimation of Au and Ag. To predict Au and Ag, the selection of finder elements was performed using multivariate statistical analysis of factor analysis and cluster analysis and linear and nonlinear regression models. In this research, it were studied 35 elements in 108 samples using statistical software Statistica 10, for forecasting and estimating the best finder elements for Au and Ag. Data were controlled based on statistical methods including identifying out-of-order elements, normalization and data standardized. Based on factor analysis and cluster analysis, the elements of the same group with Au and Ag include, Pb, Sb, As, Cd, and Zn, which can act as a finder element of Au and Ag. The purpose of the multiple regression method is to find the logical relationship between Au, Ag and finder elements. In this study, it is considered Au and Ag as dependent variables and Zn, As, Cd, Pb and Sb as independent variables. The correlation matrix shows that the correlation of Au with Zn, As, Cd and Sb is higher than 0.70 and Pb is less than 0.70. Therefore, based on linear and nonlinear regression models, Fisher test, Student t test and probability level showed that the elements of Cd, As and Sb are suitable for prediction of Au and As, Cd, Pb and Sb for prediction of Ag. In addition, non-linear regression model was found to be more suitable for prediction.