One of the possible automatiс typologies of the leukocytes in a cytometric study is proposed in this paper. Despite all the technological innovations in this area, the subjective factor strongly influences random and system errors. Some changes in the UPGMC are proposed. The modified algorithm uses a metric for determining the distance between clusters using their weights; noisy data is cleaned by the method based on the DBSCAN method; the approximation-evaluation criterion assesses the Markov moment. Cluster allocation occurs in several stages. This approach allows selecting the main groups of white blood cells for patients without serious hematopoiesis pathologies.