MapReduce-based storage and indexing for big health data

摘要

Locality Sensitive Hashing (LSH) uses randomized method to alleviate Nearest Neighbor Searchissue in high dimensional spaces. However, handling of big dataset samples for LSH algorithmbecomes difficult task because of computational complexity. So, the major aim of this work is tointroduce a new LSHalgorithmwith Hadoop MapReduce framework for enhancing proficiency ofarbitrary reads over big dataset samples. The proposed Hash index improves efficiency by reducingthe amount of accessing data for range queries by creating buckets based on hyperplanes. ALSH onMapReduce is developed, which decreases the random data access time among map andreduce functions, in addition, it enhances proficiency. Lastly, with the aim of validating the performanceof presented index for search query in MapReduce, five performance metrics such aschanging cluster size, LSH for Bucket size Balancing, the overlapped boundary of a hyperplane,Bucket creationbasedonthe configured capacity, andnon-indexed,Hashindex,andglobal indexeddataset on the HDFS configured capacity are utilized. The effect of these metrics on dataset onthe HDFS configured capacity for the period of map and reduce functions as well depicts thepre-eminence of the presented Hash index.