Challenges in Process mining
Process mining is an significantly important tool for modern organizations that need to govern non-trivial operational processes. On one side, there is an astronomical growth of event data. On the other side, processes and information need to be aligned perfectly in order to meet requirements related to compliance, efficiency, and customer service.
In spite the applicability of process mining in organizations there are still important roadblocks or challenges that are required to be addressed; these illustrate that process mining is an emerging discipline. This list might not be complete and, over time, new challenges may emerge or existing challenges may disappear due to rapid advances in process mining capabilities and rapid research and development by various organisations across the globe.
1. Finding merging and cleaning event data:
a. Data might be distributed across variety of sources. For example: certain systems may identify people using name and birth date and other might use another identifier such as Aadhaar card
b. Event data are often “object centric “rather than “process centric”. However, to monitor particular process these object centric data needs to be merged and pre-processed
c. Event data might be incomplete and in certain instances we need to extrapolate that might lead to inaccuracy
d. Event data might also contain outliers that are required to be removed. Challenge lies in defining, detecting and removing these outliers
e. Event data might have logs of varying complexity. For ex. Simple processes such as invoice processing might have different logs compared with other processes such as mortgage processing
f. Event might occur in a particular context which might introduce certain degree of errors and analysis become intractable as merging of event with contextual data might be required. For ex- rainfall might delay data speed
2. Dealing with complex event logs
a. Certain event logs may be too large to handle, whereas others might be too small to make concrete conclusions. Therefore, additional efforts might be needed to improve scalability and performance.
b. Process mining techniques need to deal with incompleteness by using an “open world assumption” — the fact that something didn’t happen does not mean that it cannot happen
c. Sometimes low level events might not give insights, hence, we might have to aggregate low level events into high level events
d. Instead of trial and error approach tools should provide detailed report on feasibility of data set for process mining
3. Dealing with process drift
a. Process might be changing while it is being analysed for example in December there might be more demand or on Friday evenings there might be less employees available etc.
b. Such changes are required to be assessed and analysed for which additional research or data might be required
4. Cross-Organizational Mining
a. There are scenarios where process flows through multiple organization and processes are needed to be mined accordingly
b. Event logs of multiple organizations need to merged and analysed.
c. New analysis techniques need to be developed for cross organizational process mining. These techniques should also consider privacy and security issues. Organizations may not want to share information for competitive reasons or due to a lack of trust. Therefore, it is important to develop privacy-preserving process mining techniques.
5. Combining Process Mining with other Types of Analysis
a. Process mining techniques can be used to learn a simulation model based on historical data. Subsequently, the simulation model can be used to provide operational support.
b. Because of the close connection between event log and model, the model can be used to replay history and one can start simulations from the current state thus providing a “fast forward button” into future based on live data
c. By combining automated process mining techniques with interactive visual analytics, it is possible to extract more insights from event data
6. Improving usability and understandability for non-experts
a. The user may have problems understanding the output or is tempted to infer incorrect conclusions. To avoid such problems, the results should be presented using a suitable representation
b. Moreover, the trustworthiness of the results should always be clearly indicated as sometimes data might not be adequate to make concrete conclusions