Logo des Repositoriums

Auflistung nach Autor:in "Chalupar, Peter"

1 - 2 von 2
  • Konferenzbeitrag
    On the Difficulties of Supervised Event Prediction based on Unbalanced Real-World Data in Multi-System Monitoring
    (Softwaretechnik-Trends Band 39, Heft 4, 2019) Schörgenhumer, Andreas; Kahlhofer, Mario; Chalupar, Peter; Mössenböck, Hanspeter; Grünbacher, Paul
    Online failure prediction of performance-critical events is an important task in fault management of software systems. In this paper, we extend our previous multi-system event prediction by analyzing its performance on unbalanced, real-world data, which represents a realistic online scenario. We train a random forest classifier with different data preprocessing configurations, including data augmentation to cope with the extreme class imbalance. The results reveal that the prediction quality of the tested multi-system model drops significantly compared to the balanced scenario. Although our supervised event prediction approach as well as different data preprocessing configurations turned out to be ineffective, we consider the insights of our work valuable for the community.
  • Konferenzbeitrag
    Using Multi-System Monitoring Time Series to Predict Performance Events
    (Softwaretechnik-Trends Band 39, Heft 3, 2019) Schörgenhumer, Andreas; Kahlhofer, Mario; Chalupar, Peter; Mössenböck, Hanspeter; Grünbacher, Paul
    The prediction of failures and other mission-critical events plays an important role in operating today’s software systems and has drawn the attention of many researchers. Event prediction is particularly challenging if multiple systems are involved. In this paper, we thus present an event prediction model which utilizes time series monitoring data from multiple software systems to predict performance events. Our approach incorporates a comprehensive, multi-system data preprocessing framework for creating various feature vector sets, which we then use to train a random forest classifier to evaluate our multi-system event prediction. Our preliminary evaluation based on data from monitoring 250 systems over a period of 20 days shows promising results.