Multiple myeloma and chronic lymphocytic leukemia are oncological diseases of the blood, which remain incurable today. The paper proposes a method for classifying blood serum samples from patients with multiple myeloma, chronic lymphocytic leukemia and healthy donors based on the analysis of their spectra in the mid-infrared (IR) range. IR spectra of blood serum were recorded using a Tensor 27 IR Fourier spectrometer in D2O solution. To analyze the obtained spectra in this work, a machine learning
algorithm was implemented – the principal component analysis. The use of the principal component analysis made it possible to significantly simplify the representation of the array of spectral data. 45 samples of blood serum were analyzed in the work. As a result of applying this approach, the studied set of samples is divided into three disjoint sets corresponding to blood serum samples of patients with multiple myeloma, chronic lymphocytic leukemia and healthy donors. Thus, the principal component method can be successfully applied to classify blood serum samples of patients with diagnoses of multiple myeloma and chronic lymphocytic leukemia. The universality of the proposed algorithm allows us to expect that in the future it is possible to apply a similar approach for other oncohematological diseases.