In the present study, classification tree induced by modified particle swarm optimization (MPSOCT) was invoked to construct a quantitative structure–activity relationship (QSAR) classification model for diverse series of antimicrobial agents against Candida albicans (CA). This may enable the screening of the antimicrobial agents in silico. In MPSOCT, modified particle swarm optimization (MPSO) was used to induce a globally optimal CT via simultaneously searching the optimal splitting parameters (i.e., splitting variables and values) and appropriate structure in tree. Based on more than hundred descriptors calculated by Material Studio 4.0 software, the classification model by MPSOCT categorized the compounds into the actives or inactives, compared with those by classification tree (CT), radial basis function network (RBFN), and partial least-squares discriminant analysis (PLS-DA). Experimental results revealed that MPSOCT offered more satisfactory classification accuracy than the other three classification algorithms. In addition, descriptors HOMO eigenvalue, Dipole_Y, S_ssCH2, S_dsCH, S_dO, AlogP98, Molecular flexibility, Wiener index, Kappa-2, Subgraph counts (2): path, Subgraph counts (3): cluster, Chi (0), Bond information content, Principal moment of inertia X, Ellipsoidal volume, Shadow_YZ, Shadow_ZX, Total molecular mass, and Atom count were found to play the most predominant roles in the antimicrobial activities against CA.