The blast furnace process is the most common form of iron ore reduction. The physical and chemical reactions in the blast furnace process are well understood on a high level of abstraction, but many more subtle inter-relationships between injected reducing agents, burden composition, heat loss in defined wall areas of the furnace, inhomogeneous burden movement, scaffolding, top gas composition, and the effect on the produced hot metal (molten iron) or slag are not totally understood. This paper details the application of data-based modeling methods: linear regression, support vector regression, and symbolic regression with genetic programming to create linear and non-linear models describing different aspects of the blast furnace process. Three variables of interest in the blast furnace process are modeled: the melting rate of the blast furnace (tons of produced hot metal per hour), the specific amount of oxygen per ton of hot metal, and the carbon content in the hot metal. The melting rate is largely determined by the qualities of the hot blast (in particular the amount of oxygen in the hot blast). Melting rate can be described accurately with linear models if data of the hot blast are available. Prediction of the required amount of oxygen per ton of hot metal and the carbon content in the hot metal is more difficult and requires non-linear models in order to achieve satisfactory accuracy.