| Accurate timestamps constitute the backbone of reliable process-mining analysis, otherwise algorithms cannot determine the true order or duration of activities. In practice, event logs frequently contain missing or erroneous timestamps, rendering large volumes of data practically unusable. This paper presents a modular Genetic Algorithm (GA) framework that treats timestamp repair as a combinatorial optimisation problem. For each missing timestamp four context-aware clusters narrow the search space to produce realistic values. A configurable GA then proposes candidate timestamp sets, evaluates them through an appropriate fitness function, and refines them via properly tuned genetic evolution process. The framework is evaluated on three BPI Challenge datasets (2015, 2017, 2019) under multiple parameter configurations varying the number of concurrent error corrections, population size, and operator control mode. The experimental results demonstrate that the proposed approach can restore missing timestamps with single-digit-second mean deviations from the ground truth, confirming that evolutionary search is a viable and effective strategy for temporal data repair in process mining under controlled evaluation conditions. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.