Actionable Pattern Mining - A Scalable Data Distribution Method Based on Information Granules
Arunkumar Bagavathi , Abhishek Tripathi , Angelina A. Tzacheva , Zbigniew W. Raś
AbstractActionable patterns are a form of recommendations that gives knowledge required by the user on actions they need to undertake to attain profit or achieve goals. Action rule is one of the methods to mine actionable patterns that are hidden in a large dataset. In the modern era of big data, organizations are collecting massive amounts of data and they keep the data updated constantly, including in major domains like financial, medical, as well as social media and IoT. Classical action rules extraction models, which analyze the data in a non-distributed fashion, lacks efficiency when processing big volumes of data and thus require a distributed approach to extract rules. Major complications of discovering action rules with distributed methods include data distribution among processing nodes in the cluster and extracting mined rule parameters like: support, confidence, utility, and coverage, representing the entire data. In this work, we focus on data distribution phase of the distributed actionable pattern mining problem. Very few works have been proposed to distribute the big data in both horizontal and vertical fashions, and extract rules/knowledge from the distributed data. We primarily focus on the vertical data splitting strategy using information granules, which give meaningful representations of complex information systems as fine grained granules. We tend to make the vertical data partitioning more logical using information granules instead of splitting the data in random.
* presented citation count is obtained through Internet information analysis and it is close to the number calculated by the Publish or Perish system.