Abstract With a large number of factors affecting the energy efficiency of buildings, the importance of analyzing the growing amount of data becomes important. The aim of this research is to check whether the use of Self-Organizing Maps will allow the indication of the relationships between building features which are considered important from the point of view of their energy performance. The research was carried out on a sample of 5040 randomly generated variants of single-family buildings with a fixed volume and location. These models were next subject to clustering, based on selected features, with the use of Self-Organizing Maps. The results prove the suitability of the method used. Grouping analysis shows the dependencies between particular parameters, confirming the importance of the U-value of partitions, the thermal mass of the building and its air-tightness for energy efficiency, while discovering unexpected relationships such as the irrelevancy of a building's orientation. In addition to expanding knowledge about the relationships between building features affecting their energy performance, it may allow optimal parameters for the given initial conditions to be found.