%0 Journal Article
%A CHENG Bo
%A LIU Chun-hong
%A MAO Wen-tao
%A WANG Meng-qing
%A ZHANG Zhi-hua
%T Joint Prediction of Multi-Workload Sequences Based on Temporal Correlation in the Cloud
%D 
%R 10.13190/j.jbupt.2019-244
%J Journal of Beijing University of Posts and Telecommunications
%P 68-75
%V 43
%N 4
%X A novel approach,joint prediction of multi-workload sequences,was proposed based on temporal correlation. Firstly,long short-term memory was used to extract the temporal feature among workload sequences for obtaining the similar workload sequences,while hierarchical clustering algorithm was used to obtain workload sequences clusters. Then,construct multi-task learning model respectively for each obtained sequence clusters,capture and utilize the shared domain knowledge among multiple workload sequences,so as to achieve joint prediction of multiple workload sequences as well as improve generalization ability and prediction accuracy of the model. Results of experiment on dataset of Google cluster trace 2011 demonstrates that the temporal feature clustering can effectively extract and utilize the global temporal feature of workload sequences,reduce the noise of original sequences and get workload sequences clusters with similar characteristics. Proposed method performs better in prediction accuracy than the state-of-the-art methods.
%U https://journal.bupt.edu.cn/EN/10.13190/j.jbupt.2019-244