Parallelizing the Bounded K-Nearest Neighbors Algorithm for Distributed Computing Systems
Arialdis Japa, Yong Shi · 2020 10th Annual Computing and Communication Workshop and Conference (CCWC) · 2020
The need for data collection and analysis in recent years has given rise to the field of Big Data. Organizations place high importance in this area because it could potentially lead to an improved quality of life for consumers and large amounts of profits. The K Nearest Neighbors (KNN) algorithm has been used to extract meaningful information from datasets. However, its performance suffers when it's applied to large datasets due to a bottleneck issue. We previously proposed a Bounded KNN algorithm, which alleviates this bottleneck and improves performance without sacrificing prediction accuracy. Although it is more efficient than the traditional KNN algorithm, the Bounded KNN is still not ideal for handling the massive amounts of data which are prevalent in today's world. In this paper, we present a parallelized algorithm which further improves performance by distributing the workload across a cluster of machines. Our experimental results show that there is some overhead time involved with distributing the workload, but when the datasets are increasingly larger, the benefits of parallelization eventually outweigh this limitation.