Analyzing process change logs provides valuable information about the evolution of process instances. This information can be used to support responsible users in planning and executing future changes. Change mining results in a change process, which represents the dependencies between process changes mined from the change log. However, when it comes to highly adaptive process settings, multiple limitations of the change process representation can be found, i.e., based on change processes it is not possible to provide answers to important analysis questions such as `How many instances have evolved in a similar way?' or `Which changes have occurred following a particular change?'. In this paper, change trees and n-gram change trees are introduced to serve as a basis to analyze changes in highly adaptive process instances. Moreover, algorithms for discovering change trees and n-gram change trees from change logs are presented. The applicability of the approach is evaluated based on a systematic comparison with change mining, a proof-of-concept implementation and by analyzing real-world data.

Top