SPOTLIGHT Data


New algorithm repairs missing data


in event logs with superior accuracy The high accuracy of the new data restoration algorithm guarantees its applications not only in current enterprises but also in future AI applications


P


rocess models and optimisation processes rely on the quality of data. Missing data can lead to models that generate


an incorrect analysis. In a new study, researchers from Pusan National University, South Korea, have developed an improved algorithm that uses correlations between existing information to restore missing data in an event log with a high degree of accuracy. Digitalisation has enabled businesses


to record their operations in event logs where each activity in a business process is recorded as data with certain attributes such as a timestamp, event name, and so on. These logs give an overview of the operations and can be used to develop process models that optimise the business process. However, the quality of the optimisation process is only as good as the data stored, and event logs with missing events lead to poor analysis and data models.


Joint effort


In a collaborative study, researchers from Pusan National University, South Korea, including Dr Sunghyun Sim and Proessor Hyerim Bae, along with Professor Ling Liu from Georgia Institute of Technology, have developed a method that can restore missing data in an event log. The study, published in IEEE Transactions on Services Computing, uses imputation methods that use correlations between available data to find missing information. “Since data is collected from multiple perspectives in numerous information systems, there is a relationship between the collected data. Starting with this point, our study suggested a method of restoring missing event values by utilising the relationship among entities in the event log, which can overcome human error or system,” said Dr Sim. In event logs, events have attributes


that are linked to other events in “single event” or “multiple event” relationships. In the former case, each attribute of an event corresponds to a unique


automationmagazine.co.uk


attribute in another event. Based on this relationship, the researchers developed a Systematic Event Imputation (SEI) method that restores a missing value by simply referring to the available value it is linked to. However, in the latter case,


where attributes have multiple correspondences, a simple matching of attributes is not possible. For such situations, a multiple event imputation (MEI) method was developed where missing events are first estimated and used to create event sequences or event chains. These sequences can be compared with an event log without missing data to restore the missing event attributes.


These imputation methods were applied simultaneously by a Bagging Recurrent Event Imputation (BREI) algorithm, bootstrap sampling and recurrent event imputation (REI) to repair the event log. On tests with real-world event logs, the researchers found that their algorithm improved restoration accuracy by 10-30%


compared to existing restoration algorithms. Moreover, it could restore almost 90% of the data accuracy even when more than half of it was missing.


Other applications Apart from optimising business processes, the researchers are optimistic that such an algorithm can be extended to other applications that rely on the quality of data. One promising avenue lies in improving the data fed to AI systems, and this method has the potential to accelerate the development of AI technologies. “It is possible to improve the performance of artificial intelligence by improving the quality of data in its learning process. The algorithm will also help prevent model malfunction by improving the quality of data it collects in real-time in a real-time environment,” said Professor Bae.


The high accuracy of the new algorithm, as well as its versatility, will ensure its widespread application in industry in the near future.


Automation | December/January 2022 9


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48