Enterprise System Response Time Prediction Using Non-stationary Function Approximations
We consider the problem of predicting response time of a large scale enterprise system using causal forecasting models. Specifically, the problem pertains to predicting potential system failure well in advance so that preventive actions can be initiated. Various influential factors are identified and their relationship with the system response time is estimated from data using non-stationary (time dependent) functional approximations. Experimental results on the prediction performance of different methods are presented and their discriminative characteristics with regard to error distribution are used to suggest a recommendation for practical implementation.
KeywordsMultivariate time series forecasting Machine learning Enterprise systems Predictive analytics
- 3.Moneta, A., Spirtes, P.: Graphical models for the identification of causal structures in multivariate time series models. In: Proceeding of Fifth International Conference on Computational Intelligence in Economics and Finance (2006)Google Scholar
- 5.Spiegel, S., Gaebler, J., Lommatzsch, A., Luca, E., Albayrak, S.: Pattern recognition and classification for multivariate time series. In: SensorKDD 2011, San Diego (2011)Google Scholar
- 6.Cheng, H., Tan, P., Gao, J., Scripps, J.: Multistep-ahead time series prediction. In: PAKDD, pp. 765–774 (2006)Google Scholar