Short-term prediction method and system for time series data in server

Abstract

The invention discloses a short-term prediction method and system for time series data in a server. The method comprises the following steps: extracting historical data; fragmenting the historical data into training data and test data; respectively training the three models by using the training data; obtaining three prediction models, comparing the symmetrical average absolute percentage error values in the three prediction models, finding out the prediction model corresponding to the minimum symmetrical average absolute percentage error value, defining the prediction model as an optimal prediction model, and storing the optimal prediction model in a database; and selecting an optimal prediction model from the database, selecting index data from the time sequence database, and calculatingthe optimal prediction model and the index data to obtain a prediction value. By means of the mode, operation and maintenance personnel can rapidly judge whether capacity increasing and capacity expanding are conducted or not according to the prediction situation, so that the efficiency of the operation and maintenance personnel can be improved, and then the competitiveness of products in the market is improved.

Description

technical field [0001] The invention relates to the field of cloud computing, in particular to a short-term prediction method and system for time series data in a server. Background technique [0002] With the rapid development of the field of cloud computing, the construction of bare metal with physical machine performance and cloud elasticity is quietly emerging in cloud computing; in order to optimize the performance of physical machines and cloud hosts in cloud computing, analysis and monitoring data for machine Performance tuning is instructive. [0003] The current monitoring data of the server mainly includes performance data such as cpu, memory, storage, network, etc. These data include time series performance data such as cpu usage rate, memory usage rate, network throughput, etc. Short-term prediction of time series data includes stationary data and non-stationary data For processing, the short-term forecasting models for non-stationary data include models such as...

AI Technical Summary

Problems solved by technology

[0004] The index data in the server includes stationary data and non-stationary data. Since there are many factors that change the stationarity of the time series data, there are periodic stable disturbances, trend changes and other random disturbances. If the same algorithm is used, it may be Causes to work well on one set of data but poorly on another

Images