Provided are data processing apparatus and methods for predicting a pattern frequency thereof. The data processing apparatus includes: a storage unit adapted to store a preset graph pattern, an occurrence frequency of the preset graph pattern, and information on actual data nodes corresponding to each node which belongs to the graph pattern; and a control unit adapted to compare the preset graph pattern with a graph pattern to be searched to perform a graph search, and when the preset graph pattern is not included in the graph to be searched, search for a node shared by two different preset graph patterns which are stored in the storage unit to perform the graph search. By this configuration, it is possible to predict a level of an intermediate result to optimize a graph query and effectively calculate sizes of various graphs patterns while making a capacity of a memory small.