Process-aware Learning

Process optimization and prediction through Process Mining and Machine Learning

© Fraunhofer IIS

Processes in companies are diverse and complex. The production of different products, inter- or intra-company transport processes or other serial event sequences within companies have one thing in common - they can be traced on the basis of a documentation of the individual process steps. In most cases, companies have process experts for the respective department who plan and control these steps using their experience and knowledge. However, with the increasing complexity and diversity of processes, efficient planning and control is becoming more difficult or even impossible for human decision makers.

In the competence pillar »Process-aware Learning«, the information documented on the data side, which is contained in the flow and execution of any process, should be integrated into Machine Learning (ML) models ­– in a way that is useful and as interpretable as possible for users. These models are used to identify important factors influencing the process, various process key figures, or anomalies in the process and, based on these, to make forecasts or recommendations for action tailored to the process flows.


Process mining generates data that can be used for machine learning

A first necessary step for the generation of ML models in the process environment is the extraction of relevant process information from the available data. For this purpose, process mining methods are used for data-driven process analysis and modeling. The information and insights gained that way must then be used sensibly as input variables for ML models in order to use these models to uncover the relationships between process steps and the other process parameters and variables recorded.

The overall goal is to explore methods that can ideally generate directly interpretable analyses and forecasts while being able to process a high complexity of the input variables and process information to be integrated. Here, the focus is not only on the use of white-box models but also on the comprehensible communication of forecasts from black-box models such as deep neural networks. A large part of the research is also devoted to the meaningful and compact formalization of the information extracted from the process data.

From process data to Machine Learning to anticipatory support systems

The basis of the research field of process analysis is data in a special form, so-called event logs. These differ from conventional data structures such as cross-sectional or time series data in that there are usually irregularly distributed data points in the form of executed activities. This makes it difficult to apply classic analysis and forecasting algorithms, but on the other hand it offers the opportunity to use methods such as process mining to extract process knowledge available on the data side.

This process knowledge traditionally contains information on the chronology of various process steps, on certain patterns in the (sub-)process flow, or on the resources used in the process, such as equipment or personnel, which have an impact on target variables and key figures of the process to be analyzed. The extracted information is then integrated into explainable ML models in order to ensure that the developed method is always comprehensible to users. Methods that fulfill these criteria and are therefore increasingly used in process-aware learning are, for example, Bayesian networks, Markov models or decision trees.

Process-aware learning in applications

The topic of analysis in process-aware learning can be process key performance indicators such as throughput times or defect rates and their influencing factors of production processes. Also complete processes and their components ("activities") as well as anomalies in processes or bottlenecks can be predicted. In this way, uncertainties in processes can be made tangible, for example, for driver assistance systems in rail transport. The data-driven forecasts and suggestions generated by ML models can serve users as a support system for process planning and control.

Analysis frameworks for the automated integration of process knowledge into process prediction models

The integration of process information into explainable machine learning models is generally associated with a high conceptual effort. Therefore, additional efforts of the competence pillar deal with the realization of an automated algorithmic generation of interpretable models for process prognosis and the thereby possible predictive support of process planning and control. In cooperation with the Ludwig-Maximilians-University Munich, methods are developed to automatically learn causal network structures from process data. Various disciplines from the field of process mining, in particular the data-driven creation of process models ("process discovery"), will be analyzed for their ability to extract causal relationships between process steps and other process parameters from the process data. The added value of such a procedure is an enormous reduction in the manual effort required to convert process data into usable analysis models while preserving the comprehensibility and interpretability of the forecasts and model-generated suggestions for process optimization.

An important step towards the application of process mining and machine learning is a qualitatively and quantitatively sufficient basis of training data. Both in the area of processing time series, cross-sectional data or text and image processing, as well as in the field of sequential data on processes, it is therefore important to research methods for the highly granular, diverse and, above all, error- and gap-free extraction or augmentation of data sets. The Data-Centric AI competence pillar is primarily dedicated to this research area. With regard to available process event logs, a great need for retrofitting or research is evident, especially in less digitized companies. For production processes, for example, it is conceivable to support the digitization of processes and thus data acquisition through cyber-physical systems (CPS) and simulation. In this way, processes can be analyzed on a data basis using the extended event log.

The competence pillar “Process-Aware Learning" is an integral part of the "Process Intelligence" group

The competence pillar “Process-Aware Learning" is an integral part of the "Process Intelligence" group of the Fraunhofer IIS Supply Chain Services workgroup. The group's researchers are dedicated to two main areas, the data-driven investigation of business processes and machine learning models for forecasting and monitoring processes.




Our focus areas within AI research

Our work at the ADA Lovelace Center is aimed at developing the following methods and procedures in nine domains of artificial intelligence from an applied perspective.

Automatisches Lernen
© Fraunhofer IIS

Automated Learning covers a large area starting with the automation of feature detection and selection for given datasets as well as model search and optimization, continuing with their automated evaluation, and ending with the adaptive adjustment of models through training data and system feedback.


Sequenzbasiertes Lernen
© Fraunhofer IIS

Sequence-based Learning concerns itself with the temporal and causal relationships found in data in applications such as language processing, event processing, biosequence analysis, or multimedia files. Observed events are used to determine the system’s current status, and to predict future conditions. This is possible both in cases where only the sequence in which the events occurred is known, and when they are labelled with exact time stamps.

Erfahrungsbasiertes Lernen
© Fraunhofer IIS

Learning from Experience refers to methods whereby a system is able to optimize itself by interacting with its environment and evaluating the feedback it receives, or dynamically adjusting to changing environmental conditions. Examples include automatic generation of models for evaluation and optimization of business processes, transport flows, or control systems for robots in industrial production.

© Fraunhofer IIS

Data-centric AI (DCAI) offers a new perspective on AI modeling that shifts the focus from model building to the curation of high-quality annotated training datasets, because in many AI projects, that is where the leverage for model performance lies. DCAI offers methods such as model-based annotation error detection, design of consistent multi-rater annotation systems for efficient data annotation, use of weak and semi-supervised learning methods to exploit unannotated data, and human-in-the-loop approaches to improve models and data.

© Fraunhofer IIS

To ensure safe and appropriate adoption of artificial intelligence in fields such as medical decision-making and quality control in manufacturing, it is crucial that the machine learning model is comprehensible to its users. An essential factor in building transparency and trust is to understand the rationale behind the model's decision making and its predictions. The ADA Lovelace Center is conducting research on methods to create comprehensible and trustworthy AI systems in the competence pillar of Trustworthy AI, contributing to human-centered AI for users in business, academia, and society.

Mathematical optimization plays a crucial role in model-based decision support, providing planning solutions in areas as diverse as logistics, energy systems, mobility, finance, and building infrastructure, to name but a few examples. The Center is expanding its already extensive expertise in a number of promising areas, in particular real-time planning and control.

© Fraunhofer IIS

The task of semantics is to describe data and data structures in a formally defined, standardized, consistent and unambiguous manner. For the purposes of Industry 4.0, numerous entities (such as sensors, products, machines, or transport systems) must be able to interpret the properties, capabilities or conditions of other entities in the value chain.

Tiny Machine Learning (TinyML) brings AI even to microcontrollers. It enables low-latency inference on edge devices that typically have only a few milliwatts of power consumption. To achieve this, Fraunhofer IIS is conducting research on multi-objective optimization for efficient design space exploration and advanced compression techniques. Furthermore, hierarchical and informed machine learning, efficient model architectures and genetic AI pipeline composition are explored in our research. We enable the intelligent products of our partners.

© Fraunhofer IIS

Hardware-aware Machine Learning (HW-aware ML) focuses on algorithms, methods and tools to design, train and deploy HW-specific ML models. This includes a wide range of techniques to increase energy efficiency and robustness against HW faults, e.g. robust training for quantized DNN models using Quantization- and Fault-aware Training, and optimized mapping and deployment to specialized (e.g. neuromorphic) hardware. At Fraunhofer IIS, we complement this with extensive research in the field of Spiking Neural Network training, optimization, and deployment.