Nowadays, there has been a meaningful increase in the use of frequent approximate subgraph (FAS) mining for different applications, for example, graph classification. However, the great amount of mined patterns is one of the fundamental drawbacks of FAS mining. This drawback has a negative effect in the computational performance of classifiers, especially in large graph databases where the number of frequent patterns could be very high. In this paper, we propose a research proposal driven to obtain FAS mining algorithms capable to compute a representative subset of patterns. The representative pattern set should be identified into the mining process improving the efficiency in time, in comparison with the time required if this identification is performed in a post-processing stage over all patterns computed by a general FAS mining algorithm.