Hits:

Indexed by:会议论文

Date of Publication:2015-01-01

Included Journals:CPCI-S

Page Number:1069-1073

Key Words:Mobile network; Prediction; Time series decomposition; K-medoids

Abstract:This paper proposes a mobile network performance warning model based on time series decomposition and K-medoids. By moving average, time series are decomposed to trend component and period-irregular component, which are processed separately then. To trend component, polynomial fitting is used to get its prediction. And to period-irregular component, K-medoids is used to eliminate irregular component, getting period index. The prediction result, as warning threshold baseline, is got by multiplying prediction of trend component and period index. Combined with historical data, warning threshold is got. The experimental results show that, this model reflects trend of performance effectively, overcomes influence of abnormal points, is fit for mobile network performance warning.