Predictive Process Mining a Systematic Literature Review

As business process complexity reaches an all-time high, the competitive and rapidly changing environments where organisations operate created the need for business processes to be continuously analysed, improved, and supported by adequate tools and techniques, which led to the conception of Process Mining (PM). Recently, Predictive Process Mining (PPM) emerged as a novel field that integrates PM with predictive techniques, such as Machine Learning, allowing the future behaviour of ongoing processes to be predicted and actions to be taken to steer them in favour of business interests. This allows PM to evolve from a reactive to a proactive tool for process enhancement, which helps reduce costs, time, and necessary resources. Nevertheless, the impact and value of PPM remain mostly unproven and the number of related studies is relatively low. A systematic review has therefore been elaborated to support future studies, based on a search carried out in three databases and according to the PRISMA statement. This review aims to identify the methods and techniques used in the implementation of PPM use cases and to understand the perceived business impact and value. In total, 411 articles were initially identified, with 24 meeting the defined criteria and being selected for discussion. From these articles, next-event, outcome, and suffix prediction have been identified as the most common use cases of PPM, in addition to the predictive techniques (e.g. XGBoost, LSTM, BERT, GAN) and optimisation methods (e.g. trace bucketing) being applied in their development. Finally, although limited, the perceived impact and value of PPM in organisations have been recorded, according to stakeholders actively using PPM in their daily work.