Data distribution method for scalable actionable pattern mining
Arunkumar Bagavathi, Varun Rao, Angelina A. Tzacheva · 2018
Action Rules are rule based systems for discovering actionable patterns hidden in a large dataset. Action Rules recommend actions that a user or a system can undertake to their advantage, or to accomplish their goal. Current Action Rules extraction methods are unable to process huge volumes of data in a reasonable time and it requires a distributed and parallel extraction methods. Limited research has been done on extracting Action Rules using a distributed scenario. Major complications of discovering Action Rules with such distributed systems are data distribution among computing nodes and calculation of major parameters of action rules. In this work, we propose few methods to handle the big data distribution among computation nodes using the Spark framework. With enhanced experiments made on datasets in transportation, medical, and business domains, we show our methods achieve almost equal valid results compared to results from classical non-distributed Action Rule discovery methods with improved run time.