For increasing accuracy of a measuring device, before carrying out measurement, implementation of graduation, which consists of the establishment of correlation dependence between indications of the measuring device and humidity of specific material, is required. For the creation of calibration dependence of the developed measuring device as the measured 10 test samples of ammonium nitrate (nitrogen) with humidity from 0% to 2.5%, with different intervals are selected. For determination of the humidity content of ammonium nitrate, it is measured the dependence of the frequency on the humidity content of the sample. For each test, several values of frequencies of the measuring generator entered in the table with a calibration frequency are received. In this work, the short review of the known methods on the creation of calibration dependence is carried out, the impossibility of application of exact methods for an objective solution because of removal need of an excessively large number of experimental data at the calibration of the measuring transducer is set and for some other reasons. The analysis of the application of some classes of functions is made for the approximation of an output characteristic. However, the described methods above should be exposed to modification and simplification. In this regard, it is offered to use the least-squares method for improvement of calibration characteristic of a measuring device in the work, the function of calibration dependence in the form of a polynom of the seventh degree is made and matrixes of a system of equations of derivatives of first order are constructed. Further, the help of applying the software environment of Mathcad calculates the made systems of equations. For this, the coefficients of the system of equations are presented in the form of a matrix, and the free terms - in the form of a vector. Further, utilization of the Gaussian method, correction factors were found, on the basis of which a mathematical model of the calibration dependence was built that allows for the optimal accuracy of the calibration of the measuring device.