AI and Machine Learning platform integration

The AI and Machine Learning Platform hosts our custom Machine Learning solution, capable of providing predictions regarding the outcome of running cases for a generic KPI (for example, it can predict whether the case will finish late or will cost too much).

While the data uploaded into IBM Process Mining describes a process by showing which steps have occurred and when, our custom Machine Learning solution adds additional information about future steps that haven't been executed yet. This additional information gives you the ability to react in advance in case of undesired process behaviors.

Furthermore, to increase the trustability of our model, explanations are provided to understand why the model is making those predictions, reasoning on the so-called influencers of the process.

Glossary

• ML: stands for Machine Learning, which is the use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyze and draw inferences from patterns in data.

• KPI: stands for Key Performance Indicator. It is a measurable value that demonstrates if a company is achieving key business objectives.

  • For example, the total lead time of the process, which should not exceed a certain threshold, to not violate an SLA.

• Prediction: the number that is provided by the Machine Learning model for each running case, according to the selected KPI.

  • For example, the expected lead time or the expected total cost for a running case

• Influencer: the explanation related to the prediction of the machine learning model that is related to an attribute of the data set and can increase or decrease the prediction.

  • For example, Closure_type=Inheritance increases the prediction

Why

The main idea behind having a Machine Learning-based engine at our disposal is to provide both predictive power and a more fine-grained overview on the whole process and on single cases, be they completed or running.

The possibility to know in advance whether a case is likely to be late and the main reasons why, gives users the ability to act on the most problematic cases. Moreover, the available level of detail makes it possible to have a much better understanding of the interactions between relevant data and process activities.