Process mining applications in healthcare is a field widely investigated in the last years. Its diffusion is driven by increasing digitalization and the availability of large quantities of clinical data, enabling hospitals, clinics, and other healthcare organizations to optimize workflows, reduce operational costs, and improve asset management. The importance of process mining lies in its potential to identify inefficiencies in processes, standardize clinical practices, support evidence-based decisions and, in general, improve the quality of care provided. The article aims to systematically review the research landscape in the field of process mining in healthcare, providing an in-depth understanding of how process mining is applied in healthcare. It contributes to the existing literature by highlighting the following aspects: the specific research topics covered (i), the extent of use of various process mining algorithms in different healthcare applications, showing their adaptability and effectiveness in specific contexts (ii), and, finally, the types and characteristics of data employed in these studies, highlighting the needs and challenges related to data in healthcare process mining (iii). Through this systematic literature review, the article can support researchers in identifying the most valuable research topic to be explored by the scientific community working on process mining in healthcare. To achieve this goal, several articles focusing on the algorithms and data employed were selected and analyzed. The final discussion highlights current research gaps, suggesting future areas of investigation, and identifies critical issues and vulnerabilities of existing process mining applications in healthcare.

Business Process Management is a domain that is composed by different research or application areas. Process Mining is one of them and it is a data-driven approach to analyze and improve processes. In Process Mining “next activity prediction” is one of the most important task. The used data are event logs i.e., a timeseries of recorded events. The event logs are mainly processed using deep learning algorithms. In this study, it was proposed the comparison of prediction performance of shallow learning algorithms with a three block Bidirectional LSTM (Bi-LSTM) architecture in predicting the next activity. The algorithms were applied on all the events logs of the BPI Challenge 2020 dataset. Results show that shallow learning algorithms outperform the three-block architecture from a minimum of 1.5 to a maximum of 6 times. This suggest that simpler algorithms may be more effective than the three-block architecture.

Process Mining merges data science and process science that allows for the analysis of recorded process data by capturing activities within event-logs. It finds more and more applications for the optimization of the production and administrative processes of private companies and public administrations. This field consists of several areas: process discovery, compliance monitoring, process improvement, and predictive process monitoring. Considering predictive process monitoring, the subarea of next activity prediction helps to obtain a prediction about the next activity performed using control flow data, event data with no attributes other than the timestamp, activity label, and case identifier. A popular approach in this subarea is to use sub-sequences of events, called prefixes and extracted with a sliding window, to predict the next activity. In the literature, several features are added to increase performance. Specifically, this article addresses the problem of predicting the next activity in predictive process monitoring, focusing on the usefulness of temporal features. While past research has explored a variety of features to improve prediction accuracy, the contribution of temporal information remains unclear. In this article it is proposed a comparative analysis of temporal features, such as differences in timestamp, time of day, and day of week, extracted for each event in a prefix. Using both k-fold cross-validation for robust benchmarking and a 75/25 split to simulate real scenarios in which new process events are predicted based on past data, it is shown that timestamp differences within the same prefix consistently outperform other temporal features. Our results are further validated by Shapley's value analysis, highlighting the importance of timestamp differences in improving the accuracy of next activity prediction.

The problem of simulating complex systems, such as production lines, industrial plants, food processing, etc., today represents an opportunity that brings with it the great advantage of limiting design costs. However, nowadays the designer, after defining and implementing the mathematical models of the studied process, may need to rebuild the whole simulation framework because he needs to modify the model of even just one subsystem. It is for this reason that in this paper, a new framework for the use of Individual Subsystem Models (ISM) for the modelling and simulation of interconnected systems has been studied and implemented. Furthermore, the study of the state of the art has revealed the lack of efficient and sufficiently general numerical algorithms, but, at the same time, it is simple to use to solve the algebraic-differential equations deriving from the ISM simulation. The proposed new approach follows the paradigm of co-simulation methods, including graph theory methods, to solve the general ISM simply and efficiently. In this approach, each subsystem is required to have its own representation independently of the other subsystems. In this way, it is always possible to replace any subsystem whenever an updated representation becomes available, making maintenance and evolution of the entire ISM flexible. Our framework calls each subsystem separately in an optimal (suboptimal) order based on the structure of the graph. Each calculated output is transferred to the input of the next subsystem in the chosen. The general procedure has been validated in the context of Linear and Time-Invariant ISMs: in these hypotheses, the stability conditions have been calculated and numerical tests have been performed which show the effectiveness of the proposed approach.