This paper presents the application of machine learning for the purpose of transformer health analysis. The state of the energy transformer is defined by the health index. Based on classic tests and traditional methods, a certain index was added to each transformer. Insulating oil and its six parameters are tested. Algorithms used to predict the state are: Multinomial Logistic Regression (MLR), Extremely Randomized Trees (ETC), K-Neighbors Classifier (KNN), Random Forests (RFC), Gradient Boosting Classifier (GBC). Finally, the results obtained by machine learning and the results of Fuzzy logic and GRNN were compared. The combination of algorithms in different ensembles led to a high level of prediction accuracy, therefore the presented model could find a place in different self-monitoring and transformer maintenance systems.