INCEPTION: Intelligent Handling of concept drift in process mining supported by Cloud Computing

In the coordinated project INCEPTION we propose to develop a proof of concept where the results of the previous R&D project BIGBISC: Bringing intelligence to business processes through soft computing in data scenarios, coordinated between the University of Santiago de Compostela (USC) and the University of Zaragoza (UNIZAR), where i) several process mining algorithms were developed for the tasks of discovery, conformance, simplification, textual description, trace clustering and change mining, and ii) several components for the processing and deployment of these algorithms in a cloud infrastructure.

In subproject 1 "Intelligent change management in process mining: explainable detection and description (XAIDrift)" the CiTIUS-Universidade de Santiago de Compostela XAIDrift team extends the development and transfer of process mining algorithms, extending it to the field of predictive process mining. In addition, emphasis is placed on the explainability component of the detected changes, so that XAIDrift is framed in the paradigm of explainable and responsible AI. This aspect is of special interest since the information related to the detected changes must be understood by non-specialized users. An adequate explanation of these changes will lead to appropriate decision making. The paradigm to be used to provide explainability to the results is natural language generation, already successfully explored within BIGBISC in other contexts of process analytics (process-to-text).

Objectives

The main objective of the project is to adapt the change management technology in process models to the needs of real business environments with large volumes of data, facilitating the detection and textual description of such changes. In addition to this general objective, the following specific objectives have been identified:

Adaptation of the change detection algorithm in process models to the results of experimentation with real customer process logs, where it is necessary to deal with the complexity of process models and trace variability.
Adaptation of process-to-text technology for natural language description of changes in the model representing the observed behavior of a process.
Design and development of a service-oriented infrastructure that integrates the algorithms for the detection and description of process change.
Licensing and transfer of the proposal oriented mainly to companies that already have process mining technology and therefore have the need to perform a more detailed analysis of process behavior.

In the coordinated project INCEPTION we propose to develop a proof of concept where the results of the previous R&D project BIGBISC: Bringing intelligence to business processes through soft computing in data scenarios, coordinated between the University of Santiago de Compostela (USC) and the University of Zaragoza (UNIZAR), where i) several process mining algorithms were developed for the tasks of discovery, conformance, simplification, textual description, trace clustering and change mining, and ii) several components for the processing and deployment of these algorithms in a cloud infrastructure. In subproject 1 "Intelligent change management in process mining: explainable detection and description (XAIDrift)" the CiTIUS-Universidade de Santiago de Compostela XAIDrift team extends the development and transfer of process mining algorithms, extending it to the field of predictive process mining. In addition, emphasis is placed on the explainability component of the detected changes, so that XAIDrift is framed in the paradigm of explainable and responsible AI. This aspect is of special interest since the information related to the detected changes must be understood by non-specialized users. An adequate explanation of these changes will lead to appropriate decision making. The paradigm to be used to provide explainability to the results is natural language generation, already successfully explored within BIGBISC in other contexts of process analytics (process-to-text). The main objective of the project is to adapt the change management technology in process models to the needs of real business environments with large volumes of data, facilitating the detection and textual description of such changes. In addition to this general objective, the following specific objectives have been identified:<ul><li>Adaptation of the change detection algorithm in process models to the results of experimentation with real customer process logs, where it is necessary to deal with the complexity of process models and trace variability.</li><li>Adaptation of process-to-text technology for natural language description of changes in the model representing the observed behavior of a process.</li><li>Design and development of a service-oriented infrastructure that integrates the algorithms for the detection and description of process change.</li><li>Licensing and transfer of the proposal oriented mainly to companies that already have process mining technology and therefore have the need to perform a more detailed analysis of process behavior.</li></ul> - Alberto José Bugarín Diz, Manuel Lama Penín - Senén Barro Ameneiro, Juan Carlos Vidal Aguiar, Manuel Mucientes Molina, Víctor José Gallego Fontenla, Yago Fontenla Seco