Discovery of Action Rules at Lowest Cost in Spark

Abstract

Action Rules or Actionable patterns is a type of rule-based approach in data mining that recommends to a user specific actions, in order to achieve a desired result or goal. The amount of data in the world is growing at an exponential rate, doubling almost every two years. Distributed computing platforms like Hadoop and Spark, have eased the computation of this high velocity data. Leveraging these cutting-edge technologies in the field of Data Mining to process huge volumes of data can improve the performance and allow user to gain insights from large datasets with quick turnaround time. In this paper, we present an approach for discovering low cost actionable patterns, and provide actionable recommendations. We adapt this algorithm to distributed environment using Apache Spark framework. We evaluate the performance of the algorithm with two datasets in transportation and medical domain.