Title: An incremental clustering pattern sequence-based short-term load prediction for cloud computing
Authors: Dayu Xu; Xuyao Zhang
Addresses: Zhejiang Provincial Key Laboratory of Forestry Intelligent Monitoring and Information Technology Research, Zhejiang A&F University, Hangzhou, Zhejiang, P.O. Box 311300, China ' School of Economics & Management, Zhejiang A&F University, Hangzhou, Zhejiang, P.O. Box 311300, China
Abstract: Short-term load prediction is a significant cost-optimal resource allocation and energy saving approach for a cloud computing environment. Traditional linear or nonlinear prediction models that forecast future load directly from historical information appear less effective. Load classification before prediction is necessary to improve prediction accuracy. In this paper, a novel clustering algorithm and prediction approach is proposed to forecast future load for cloud computing data centres. First, an incremental kernel k-means clustering based data clustering method is adopted to classify the continuously coming cloud load. Secondly, Hausdorff distance based similarity computation method is then used to identify the most appropriate cluster that possesses the maximum likelihood for current load. With the data from this cluster, a fast neural network is used to forecast future load. Experimental results show that our approach is more efficient and outperforms other approaches reported in previous works.
Keywords: cloud computing; load prediction; incremental kernel k-means clustering; neural networks; pattern sequences; short-term predictions; resource allocation; energy saving; energy consumption; Hausdorff distance.
International Journal of Grid and Utility Computing, 2016 Vol.7 No.4, pp.304 - 312
Received: 21 Sep 2015
Accepted: 25 Oct 2015
Published online: 14 Dec 2016 *