Query Operators for Transactional Data: Detecting Similar and Periodic Transactions

Abstract

Abstract Pattern detection for revealing the patterns of users’ behavior is an important analysis-assisting tool toward the understanding and prediction of their attitudes, manners, activities and habits. In this paper, two novel query operators applied to transactional data are introduced to ease the query processing, strengthening query capabilities and revealing valuable patterns for data analysis and mining. The operators are named as PeriodicTransactions and SimilarTransactions, and as their names imply, they measure periodicity and similarity, respectively, in a set of transactions. The operators are formally defined and the corresponding algorithms are also provided. To show the expediency of the operators, the proposed algorithms are implemented and a set of experiments were conducted with real data from the Ethereum blockchain. The results show the feasibility and usefulness of the proposal for identifying these patterns that help to understand user behavior and reveal a rich interaction between senders and recipients, where periodic and similar transactions occur.