Process mining is the research field studying how to automatically analyze logs coming from informations systems in order to obtain a deep knowledge of the underneath business process aimed at monitoring its performance, predicting future situations and proactively acting on the process itself. While this field has been very active in the last decades, proposed techniques can be only applied when actions and activities inside logs are very high-level (e.g., a worker switched on a specific machine). In the last few years though, it has been noticed how vast majority of processes produce a lot of low level data directly coming from sensors monitoring the so called cyber-physical environments where the business processes are executed. Notable examples of cyber-physical environments are represented by smart environments and smart homes in particular, which are increasingly equipped with sensors and actuators. In this scenario, it has been argued that, human habits can be considered as example of processes that are carried on by persons during their daily lives.
Unfortunately, sensor logs produced by such environments are far from being optimal for currently available process mining techniques. For example, a common prerequisite of process mining techniques is to have a trace log explicitly segmented in traces (process instances). This assumption is usually not met by sensor logs as labeling is generally an expensive task to be performed by humans. Additionally, there is no one-to-one correspondence between sensor measurements and actions performed by a person. Techniques from the machine learning field can be used to aggregate sensor measurements in order to recognize actions, but they are very sensor-specific. Finally, discovered processes, in the case of human habits, can be so unregulated to resemble a so called "spaghetti" process, i.e., a process which is hard to visualize.
The aim of this project is to find valuable solutions to overcome these challenges.
The research envisioned in this project proposal is highly innovative as it is intended to contribute to a brand new research field, i.e., IoT-enhanced process mining in a context, the one of smart homes and smart environments in general, which is living an explosion in these years thanks to the now widespread home assistants (e.g., Google Home, Alexa, Apple Homekit) that boosts the deployment of devices and sensors inside houses. Potential contributions to the state of the art may be the following:
C1. Definition of a standard representation format for IoT-enhanced process data. As discussed, the current standard format XES for process logs has not been adapted to also represent data coming from IoT sensors. As the XES format allows for extensions, an extension can be proposed to the community that will allow to have standard datasets to be used as benchmarks.
C2. Creation of freely available synthetic datasets for the community. On of the tasks in WP1 is the development of a dataset generation tool with a certain realism level. Datasets produced through this tool will be made available to the community as benchmarks that will be also used to organize specific workshop in international conferences, thus triggering the development of a community.
C3. Integrated visualization tools. An important goal of business process management (BPM) is the visual analysis of processes. Even when precise measures are not available, the visual inspection of logs and processes can provide useful insights to analysts. Unfortunately, in IoT enhanced process, data to be displayed is vastly heterogeneous and, in particular, no scientific result is available mixing up visualization techniques from the area of business processes with data coming from streaming IoT sensors. Such tools are the goal of WP2.
C4. IoT-enhanced ready process discovery techniques. In the area of smart spaces, the principal investigator was one of the first to work on the integration of IoT data and business processes, also inventing the term "Habit Mining" for this kind of task. As other techniques in the area anyway, proposed techniques work well in very controlled conditions. One of the most important contribution to the community will be the development of techniques that work in any kind of smart space will little effort, thus complementing the manual rule definition techniques nowadays provided by tools like IFTTT, which are now integrated in home assistants.
The above identified challenges will be tackled through the research program developed for the project (see the previous section). It is worth to notice that the Principal Investigator (PI) has already obtained remarkable scientific results in the area. Obtained results will be submitted to A, A* conferences in the area of process mining and ubiquitous computing.
Notably, the results of the research, even if originally achieved in the scenario of smart homes, can be extended and generalized in similar scenarios, such as
- smart factories (Industry 4.0) in which to understand habits of workers can support the optimisation of procedures as well as the safety of the workers;
- smart hospitals, in which the understanding of the habits of doctors and nurses can support a greater efficacy and efficiency of their acting;
- smart environments with crowded persons (e.g., museums, restaurants, etc.) in which to understand the habits of persons can help in providing customised services adding satisfaction to the customers' experience, but also guaranteeing safety in the overall space (as the COVID-19 emergency and the need of social distancing while preserving sociality as created completely new needs and requirements for management of public spaces).
Therefore the impacts of the proposed research are wide and potentially applied in multidisciplinary fields.