SARGS method for distributed actionable pattern mining using spark
Arunkumar Bagavathi, Pranava Mummoju, Katarzyna A. Tarnowska, Angelina A. Tzacheva, Zbigniew W. Raś · 2017
Actionability is a mode of revealing actionable knowledge in the form of Action Rules from large datasets. Action rule imparts in the form of recommendations as how a data object can change from one value to another more desirable value. The towering production of data in the recent years, due to increased usage of web, social media and IoT, has led to the age of big data. Also, abundant usage of cloud storages and cloud based services causes the data to be spread around the globe. This requires more time and space for a single computer to cope with such widespread data. Ecosystems like Hadoop MapReduce, Spark have been introduced to store, process and retrieve back the data efficiently in a distributed fashion. Data mining finds substantial improvements over such distributed frameworks to process huge volume of data and acquire knowledge from them in a short span of time. In this paper, we present an approach SARGS: Specific Action Rule discovery based on Grabbing Strategy, to build more specific Action Rules using Apache Spark framework and evaluate the results with our previous Hadoop MapReduce system (MR-Random Forest Algorithm for Distributed Action Rules Discovery). Also, we propose a novel approach to distribute data in a distributed environment to get more optimal Action Rules and upgraded ARoGS algorithm to get more specific Action Rules.