Application of machine learning methods to improve business processes

Over time, business processes are becoming more complex. The aim of the paper is to show that machine learning methods can be successfully used to understand and improve business processes. This paper focuses on the analysis of two types of data, namely aggregated data and sequence data. Aggregated data is the result of transformations over raw data, focused on concepts that are not obvious in the source data. This aggregation is similar to the construction of functions used in the field of machine learning. The construction of functions consists in converting the original representation into a new one, usually in a more compact form, which covers most (or most relevant) initial characteristics. In this paper, aggregated data are variables that arise as a result of creating the concept of the complexity of the process. These aggregated data are used to develop the logistics of homogeneous clusters. This means that the elements in different clusters are different, in terms of the complexity of routing. The development of homogeneous clusters for this process is important in connection with the induction of predictive models. Routing in the process can be predicted using logistical clusters. These sequences describe the sequence of actions over time during the execution of the process. They are written to the process log during the execution of the process steps. Due to exceptions, absence or incompleteness of recording and errors, the data can be noisy. The use of sequence data is a tool to achieve the goal, which is to obtain a model explaining the recorded events. In situations without noise and with sufficient information, a method is provided for constructing a process model from the process log. Machine learning methods are particularly useful when detecting a process model from noisy sequence data. Such a model can be further analyzed and ultimately improved.