Improvement of Power Spectrum-area Fractal Model using linear discriminant analysis in frequency domain of geochemical data

The methods and techniques for separation ofgeochemical anomaly, background and noise havebeen the issue of research for many years. Severalpeople have applied the Power Spectrum-area (SA)Fractal model to identify geochemicalanomalies, background and noises. The purpose ofS-A method is to divide the power spectrum (PS)values into components characterized by similarscaling properties. The high frequencies in surfacegeochemical distribution map are related to surfacemineral deposits and geochemical noises. Very lowfrequencies are related to background values andvery deep mineral deposits.S-A fractal method can be used to identify the deepand blind mineral deposits. Very low frequencies insurface geochemical data that have directrelationship to high PSs can be separated using S-Afractal method. In this study, investigations havebeen conducted on Dalli Cu-Au porphyry depositthat is located in the central province of Iran. Threeranges and two cutoff values for PS have beenidentified for Cu using S-A Model. This paperintroduces the combined approach of S-A fractalmethod and linear discriminant analysis (LDA) infrequency domain of geochemical data to improvethese threshold values. The LDA is a supervisedclassification method, which investigates arelationship between features space and adependent categorical class using a set of trainingdata and be able to verify the results of clusteranalysis. In this study, LDA method has beenperformed on the results of the S-A method then theconfusion matrix has been formed and correctclassification rate indexes (CCR) have beencalculated in different threshold values. Then theresults were compared together and the optimumthreshold value between low and moderate PSvalues were determined. The results of S-A fractalmethod have been properly improved using theproposed combined method of S-A fractal andLDA.

کلیدواژه ها:

Power Spectrum-area Fractal ، linear discriminant analysis (LDA) ، anomaly threshold value ، frequency domain of geochemical data ، 2Dimensional Fourier Transformation