Fig. 3.

Contribution rates of input items for each feature vector. The items with the six highest contribution rates are shown for each feature vector, with contribution rates defined as the absolute values of the means of partial differentials for each sample. Partial differentials were calculated by defining the encoder outputs, i.e., the coordinates of the x, y and z axes, as the objective variables, and defining the input items as explanatory variables. This made it possible to qualitatively evaluate the relationship between the values of the input items and the coordinates in the three-dimensional feature space. Anti-dsDNA antibody, anti-double-stranded deoxyribonucleic acid antibody