Event suffix and remaining time prediction are sequence to sequence learning tasks.They have wide applications in different areas such as economics, digital health, business process management and IT infrastructure monitoring.Timestamped event sequences contain ordered events which carry at least two attributes: the event's label and its timestamp.Suffix and remaining time prediction are about obtaining the most likely continuation of event labels and the remaining time until the sequence finishes, respectively.Recent deep learning-based works for such predictions are prone to potentially large prediction errors because of closed-loop training (i.e., the next event is conditioned on the ground truth of previous events) and open-loop inference (i.e., the next event is conditioned on previously predicted events).In this work, we propose an encoder-decoder architecture for open-loop training to advance the suffix and remaining time prediction of event sequences.To capture the joint temporal dynamics of events, we harness the power of adversarial learning techniques to boost prediction performance.We consider four real-life datasets and three baselines in our experiments.The results show improvements up to four times compared to the state of the art in suffix and remaining time prediction of event sequences, specifically in the realm of business process executions.We also show that the obtained improvements of adversarial training are superior compared to standard training under the same experimental setup.

It is common for business processes to exhibit a high degree of internal heterogeneity, in the sense that the executions of the process differ widely from each other due to contextual factors, human factors, or deliberate business decisions. For example, a quote-to-cash process in a multinational company is typically executed differently across different countries or even across different regions in the same country. Similarly, an insurance claims handling process might be executed differently across different claims handling centers or across multiple teams within the same claims handling center. A subset of executions of a business process that can be distinguished from others based on a given predicate (e.g. the executions of a process in a given country) is called a process variant. Understanding differences between process variants helps analysts and managers to make informed decisions as to how to standardize or otherwise improve a business process, for example by helping them find out what makes it that a given variant exhibits a higher performance than another one. Process variant analysis is a family of techniques to analyze event logs produced during the execution of a process, in order to identify and explain the differences between two or more process variants. A wide range of methods for process variant analysis have been proposed in the past decade. However, due to the interdisciplinary nature of this field, the proposed methods and the types of differences they can identify vary widely, and there is a lack of a unifying view of the field. To close this gap, this article presents a systematic literature review of methods for process variant analysis. The identified studies are classified according to their inputs, outputs, analysis purpose, underpinning algorithms, and extra-functional characteristics. The paper closes with a broad classification of approaches into three categories based on the paradigm they employ to compare multiple process variants.

The alignment of observed and modeled behavior is an essential element for organizations, since it opens the door for conformance checking and enhancement of processes. The state-of-the-art technique for computing alignments has exponential time and space complexity, hindering its applicability for medium and large instances. In this article, a novel approach is presented to tackle the challenge of computing an alignment for large-problem instances that correspond to well-formed process models. Given an observed trace, first it uses a novel replay technique to find an initial candidate trace in the model. Then a local search framework is applied to try to improve the alignment until no further improvement is possible. The implementation of the presented technique reveals a magnificent reduction both in computation time and in memory usage. Moreover, although the proposed technique does not guarantee the derivation of an alignment with minimal cost, the experiments show that in practice the quality of the obtained solutions is close to optimal.