The pipeline method FPE is shown in Figure 1, which consists of five steps, each of which is implemented by a module. First, each sequence is preprocessed by the preprocessing module. Second, each sequence is divided into blocks of the same size by the sequence blocking module. Then, a modified PrefixSpan [21] algorithm is used to discover the maximal frequent patterns in each block, which is done by the modified PrefixSpan module. The entropy of each block is calculated with the probabilities of discovered patterns in the block, which is finished by the entropy calculation module. For each sequence, the entropy values of all blocks form the dimensional components of its final representation vector. Finally, the similarity calculation module computes the similarity between any two sequences using their vectors obtained above. We describe each step in detail in the following subsections