This study proposes a multi-level classifier to predict heart necrosis or risk based only on the descriptive parameters of a new laboratory animal when solving a pharmacology task. The operation of the multi-level classifier is based on data mining tasks and algorithms. The construction of the multi-level classifier uses intelligent data analysis techniques like classification, clustering and prediction. The classification process includes splitting of the data set, feature selection based on their information, growing of a decision tree using the C4.5 algorithm and induction of a conditional rule set. The clustering is based on finding groups of similar objects in heart contraction power data. The prediction is carried out using only descriptive parameters that are projected onto obtained decision tree determining a connection between descriptive parameters of an animal and the class obtained in clustering. Based on the acquired results a rule set is selected from database that is the basis for finding occurrence frequency statistics used in the calculation of the potential risk