Reliable frequent itemsets mining with actor based Apriori algorithm

This paper presents an actor based Apriori algorithm enhanced with fault tolerance mechanism. All phases of the algorithm including candidate generation and support counting operations are performed by asynchronous actors. When an error occurs during the execution of the algorithm, calculations are interrupted locally for specific actors. The actor state is restored from the snapshot and the operations that caused the failure are either repeated or skipped. Other actors progress with their current tasks. The algorithm can be executed in parallel and distributed environments. Proposed enhancements have been successfully implemented using JAVA and Akka library. This paper discusses the results of the performance of actor-based Apriori algorithm against different datasets. The presented approach has been illustrated with many experiments and measurements performed using multiprocessor and multithreaded computer.