This study is to apply the Data Mining concept to process technological data, which are received from a field level of an industrial control system of a steel billet heating in a continuous furnace of a rolling-mill shop. The main scope of the study is to find a regression dependence of the steel billet temperature on a history of its heating in such a furnace. This paper describes: 1) the principles of data warehouse implementation to store values of signals obtained from a steel billets tracking system, 2) the chosen features (input variables, regressors) to solve the problem under consideration, such as temperature in all furnace zones and time of a steel billet heating in each of them, 3) the applied algorithms of data pre-processing and exploratory analysis. Then the results of research on various machine learning algorithms application to find the required dependence are presented. The best quality of approximation is achieved using a model, which is based on classification and regression trees. They are trained with the help of an extreme gradient boosting algorithm modification-DART. Obtained prediction accuracy is ~9°C.