For many years, Heat Transfer specialists have collected data in the lab to enable the calculation of heat transfer coefficients and pressure drops in various systems. Empirical and semi-empirical correlations have typically been developed to estimate such quantities over a wider range of operating and design conditions and to enable the design of cooling and heating devices. However, with this approach, a significant amount of prior knowledge is needed to select the input variables used in the regression. Unfortunately, when it comes two phase systems, this prior knowledge is very difficult to acquire, owing the complexity of the underlying phenomena. Machine Learning algorithms offer a novel and promising approach, as they leverage from large data-sets collected in the lab, and attempt to predict key quantities with a more limited knowledge of the system itself. Nevertheless, there are still several questions open on the use of these algorithms to explain heat transfer data, in particular around model overfitting and extrapolability of the estimations. As a result, it is important to further investigate these methods and develop guidelines to formulate ML models correctly, that is, to appropriately define training and testing set, to avoid model overfitting and to test extrapolation capabilities. This article aims to (i) showcase the benefits and limitations of the use of Machine Learning algorithms (namely Random Forest and Deep Neural Networks) in regressing heat transfer data; (ii) compare the results to semi-empirical correlations; (iii) establish the relative importance of each dimensional variable in the explanation of the heat transfer coefficient; (iv) assess the difference in the regression performance of using non-dimensional numbers commonly used in the design instead of dimensional variables and (v) testing extrapolability to unseen data. A case study showing the estimation of the condensation heat transfer coefficients in micro-fin tubes using a database of 4,333 data points is presented to illustrate the above procedures and objectives.