Moderní metody získávání znalostí z procesů
In this thesis, we propose an approach for modelling hidden and unknown processes and subprocesses in the example of a seaport logistics area. Having the underlying process model makes it possible to exploit more advanced algorithms since deviations and main paths are becoming visible and better controlled. The obtained model is the foundation for the core research of this work and will be enriched with key performing indicators and their forecast by applying advanced process mining, statistics, and machine learning techniques. The main difference of the approach is that we take as a target variable not any specific value, but the object - a process variant or a process type with a set of parameters. Bottleneck analysis, from one side, and predictive analysis, on the other hand, are enforced with context-aware information, especially with these additional objective process attributes. Furthermore, the support of the descriptive ("As is") current process model with certain notation and the integration with relevant bottleneck and predictive methods compromise the advantages of the approach. The work primarily focuses on the design of algorithms and methods for supporting logistics data analysis. However, it can be adjusted and applied to other areas accordingly, which makes the approach flexible and versatile. The result of the work is the framework for unstructured process modelling and the key process parameters predictive method. This analysis of processes with their attributes might be used for decision-making systems and process maps in future.
Process mining, Statistics, Predictive models, Data mining, Nautical process, Logistics process, Process modelling, Bottleneck analysis, Process maps, Context awareness