Abstract
Multi-track data are multi-set sequences that are suitable for representing time series data, such as multi-sensor data, polyphonic music data and traffic data. The permuted pattern matching problem aims to determine the occurrences of multi-track patterns in multi-track text by allowing the order of the pattern tracks to be permuted. In this study, we address permuted pattern matching by proposing a new data structure called a filtering multi-set tree (FILM tree). The FILM tree is a complete binary tree based on a spectral Bloom filter (SBF) with hash functions. This data structure is very simple but powerful, and it can be applied to both exact and approximate matching problems. We present experimental results that demonstrate the efficiency of our FILM tree-based approach.
