An integral part of data mining process is the model building process, based on mined data and subjected for visualization/ interpretation, statistical methods that involve statistical modeling like regression do involve model building based on training data and its evaluation based on test data.
The SEM model is a measurement model that includes regression analysis, whose purpose is to predict while the SEM infers a possible causality, so it does not qualify within the
DM is about finding interesting patterns and knowledge from data ,so that i believe any technique achive that goal could be considred as DM technique ..I think there is no boundries.