This thesis focuses on adaptive analysis of several performance monitoring metrics and alerting on changes of their values. Operators of monitoring systems typically suffer from difficulties in setting proper alerts, e.g. by defining a precise value of threshold and later updating it as metric values change. The student will study existing approaches to monitoring and analytics and get acquainted with metrics used in systems in GoodData. Then he or she will propose a system that will adapt and raise only a limited number of alerts. So a model of metric value behavior should be incorporated. The model can be initially based on operator’s feedback based on his or her experience, but an automatic solution is preferred. The result of this thesis will be a survey of existing approaches to analyze trends in metrics, description of the proposed algorithms and results of performance evaluation on data provided by GoodData. A software prototype of implemented algorithms will also be attached.