It is possible to determine the morbidity of a disease by learning with a neural network using data on the expression level of a biomarker, and to extract a biomarker characteristic of the disease using a neural network. A sample model in which the expression level of each of a plurality of types of biomarkers is recorded for each individual is acquired and machine learning is performed using training data to generate a learned model that can determine the prevalence of a disease obtained in advance. Then, for this learned model, a plurality of sample data with label information affected by a disease are input and calculated, and a plurality of biologics obtained by the learned model by calculating the morbidity for each sample data. The importance level of each marker feature is digitized, and a predetermined number of biomarkers are extracted as characteristic biomarkers related to the disease based on the digitized importance level of all sample data for each biomarker. [Selection] Figure 2
展开▼
