Machine learning is a branch of artificial intelligence that describes the “learning” of a digital system on the basis of real data. An algorithm examines large amounts of data according to certain patterns or rules. Based on these findings, IT systems are able to find solutions for specific problems. This information can be used, for example, to solve related problems or to analyze unknown data. This technique is based on self-learning algorithms, which are adapted or changed during the learning process. In process mining, Machine Learning is used for root cause analyses.
How exactly does machine learning work?
Let’s start from the beginning—with us, the people. The logical prerequisite for machine learning is the existence of data. In order to generate data, some kind of human action is necessary. Take, for instance, business processes. The IT systems used here log process cycles, such as the activities of employees. But an automated data collection process was originally set up by a human. As soon as data is available, a person still has to tell the system how to analyze it. The system therefore needs rules or problems so that the algorithms “know” what they have to do to analyze the data. The human therefore supplies the input—the generated data—and defines the framework conditions for the system.
What can machine learning do?
As soon as certain patterns are identified, the system records the results in an appropriate form, for example as rules, graphics, in text form, or as key figures. These are then formulated in such a way that they can be interpreted by humans in the following ways:
- Identify, extract, and summarize data.
- Make predictions.
- Calculate occurrence probabilities of events.
- Adapt the system independently to different circumstances and developments.
- Use data to optimize processes.
Machine learning and process mining.
Large amounts of data, specifically process data, are also examined in process mining. This is why machine learning is suitable for finding out why a certain process variant was developed. This type of analysis is also known as automated root cause analysis. The algorithm examines the process data according to certain patterns or rules in order to explain process weaknesses or undesired process deviations, for example. These are then formulated as rules that provide information about the cause and extent of the vulnerability. At the same time, the rules can be used to make predictions, for example to assess the risk of bottlenecks.