Volume 21, No. 2, 2024
An Integrated Local And Global Post Filtering And Probabilistic Kernel Density Approach On Different Skyline Databases
V.Narendra Babu and A.Suresh Babu
Abstract
Parallel processing is increasingly integral to modern computing systems, particularly in managing the complexity of daily processing tasks. Skyline query processing, a critical technique in multi-criteria decision-making, is gaining prominence in large-scale computing environments. Traditional approaches like MapReduce have struggled with scalability due to the inclusion of unnecessary information, leading to inefficiencies. To address this, we propose a novel MapReduce-based framework, MapReduce - Pre and Post Filter-based Skyline Computation (MR-PPFS), which strategically limits the number of search points, enhancing both computational memory and efficiency. The growing size of spatial datasets further complicates the skyline computation, making segmentation and spatial pattern prediction increasingly challenging. Our approach overcomes these issues by implementing a hybrid multi-level parallel skyline computational model using pre-local and post-global filtering techniques on large spatial databases. This method leverages a k-nearest neighbor (k-NN) mapping class within the Hadoop framework, significantly improving the accuracy and relevance of skyline computations. Experimental results confirm that the proposed MR-PPFS model outperforms existing methods, providing a robust solution for handling large-scale spatial data with enhanced precision and reduced computational overhead.
Pages: 207-222
Keywords: Map reduce Skyline processing, Spatial data minings Hadoop.