A nutrient solution recipe analysis method was developed based on biplot method. The difference of 〖JP2〗five recipes containing seven nutrient elements, including NO3N, K+, Ca2+, Mg2+, PO4P,〖JP〗 NH4N, SO4S was discussed. The result revealed NO3N, K+, Ca2+ got a better performance than other elements in terms of the difference and content. Meanwhile, Yamazaki nutrient solution recipe and Cornell nutrient solution recipe independently hold a high correlation, respectively. A unified modeling samples collection training support vector machine (SVM) model was established for the detection of NO3N, K+ and Ca2+. Compared with the prediction model without difference analysis as input, the mean relative error of NO3N, K+, Ca2+ was decreased from 7.66%, 11.88% and 11.55% to 6.41%, 6.14% and 10.20%, respectively, in the Yamazaki recipe environment. While in the Cornell formula environment, the mean relative error was reduced by 1.79 percentage points, 2.98 percentage points and 1.13 percentage points, respectively.