Residential choice behaviour is a complex process underpinned by both housing market restrictions and individual preferences, which are partly conscious and partly tacit knowledge. Due to several limitations, common survey methods cannot sufficiently tap into such tacit knowledge. Thus, this paper introduces an advanced knowledge elicitation process called SilverKnETs and combines it with data mining using random forests to elicit and operationalize this type of knowledge. For the application case of the city of Leipzig, Germany, our findings indicate that rent, location and type of housing form the three predictors strongly influencing the decision making in residential choices. Other explanatory variables appear to have a much lower influence. Random forests have proven to be a promising tool for the prediction of residential choices, although the design and scope of the study govern the explanatory power of these models.