Cloud Based K-Means Clustering Running as a MapReduce Job for Big Data Healthcare Analytics Using Apache Mahout

Increase in data volume and need for analytics has led towards innovation of big data. To speed up the query responses models like NoSQL has emerged. Virtualized platforms using commodity hardware and implementing Hadoop on it helps small and midsized companies use cloud environment. This will help organizations to decrease the cost for data processing and analytics. As health care generating volumes and variety of data it is required to build parallel algorithms that can support petabytes of data using hadoop and MapReduce parallel processing. K-means clustering is one of the methods for parallel algorithm. In order to build an accurate system large data sets need to be considered. Memory requirement increases with large data sets and algorithms become slow. Mahout scalable algorithms developed works better with huge data sets and improve the performance of the system. Mahout is an open source and can be used to solve problems arising with huge data sets. This paper proposes cloud based K-means clustering running as a MapReduce job. We use health care data on cloud for clustering. We then compare the results with various measures to conclude the best measure to find number of vectors in a given cluster.