A Deep Learning Approach for Repairing Missing Activity Labels in Event Logs for Process Mining

Lu, Yang; Chen, Qifan; Poon, Simon K.

doi:10.3390/info13050234

Cited by 5 publications

(7 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [19], a novel classification event imputation method is proposed, which can recover missing categorical events by learning structural features observed in the event log. In [20], an LSTM-based prediction model that uses the prefix and suffix sequences of events with missing activity labels as input to predict the missing labels is proposed, demonstrating high repair capability. In [21], the BERT4Log model and weak behavior profile theory, combined with a multi-layer multi-head attention mechanism is introduced, for interpretable repair of low-quality event logs.…”

Section: A Attribute-level Repairmentioning

confidence: 99%

See 1 more Smart Citation

Log Clustering-based Method for Repairing Missing Traces with Context Probability Information

Fang,

2024

ijacsa

View full text Add to dashboard Cite

In real business processes, low quality event logs due to outliers and missing values tend to degrade the performance of process mining related algorithms, which in turn affects the correct execution of decisions. In order to repair the missing values in event logs under the condition that the reference model of the process system is unknown, this paper proposes a method that can repair consecutive missing values. First, the event logs are divided according to the integrity of the trace, and then the cluster algorithm is applied to complete logs to generate homogeneous trace clusters. Then match the missing trace to the most similar sub log, generate candidate sequences according to the context of the missing part, calculate the context probability of each candidate sequence, and select the one with the highest probability as the repair result. When the number of missing items in the trace is 1, our method has the highest repair accuracy of 97.5 percent in the Small log and 93.3 percent in the real event logs bpic20. Finally, the feasibility of this method is verified on four event logs with different missing ratios and has certain advantages compared with existing methods.

show abstract

Section: A Attribute-level Repairmentioning

confidence: 99%

“…F1: Deterministic repair for known anomalies or missing attributes, where the position of the anomaly or missing attribute is clearly identified. Attribute-level repair [13], [17], [19], [20] ✓ ✓…”

Section: Summary Of Existing Workmentioning

confidence: 99%

Log Clustering-based Method for Repairing Missing Traces with Context Probability Information

Fang,

2024

ijacsa

View full text Add to dashboard Cite

show abstract

“…This review identifies research gaps and proposes a research agenda for its application in various business contexts. Chen et al [13] observed that process mining bridges process modelling and data mining. To propose an LSTM-based model to repair missing activity labels in event logs, outperforming existing methods.…”

Section: Related Workmentioning

confidence: 99%

PMiner: Process Mining using Deep Autoencoder for Anomaly Detection and Reconstruction of Business Processes

Chinnaiah,

Veerabhadram,

Aavula

et al. 2024

Int. j. electr. comput. eng. syst. (Online)

View full text Add to dashboard Cite

We proposed a deep learning-based process mining framework known as PMiner for automatic detection of anomalies in business processes. Since there are thousands of business processes in real-time applications such as e-commerce, in the presence of concurrency, they are prone to exhibit anomalies. Such anomalies if not detected and rectified, cause severe damage to businesses in the long run. Our Artificial Intelligence (AI) enabled framework PMiner takes business process event longs as input and detects anomalies using a deep autoencoder. The framework exploits a deep autoencoder technique which is well-known for Its ability to discriminate anomalies. We proposed an algorithm known as Intelligent Business Process Anomaly Detector (IBPAD) to realize the framework. This algorithm learns from historical data and performs encoding and decoding procedures to detect business process anomalies automatically. Our empirical results using the BPI Challenge dataset, released by the IEEE Task Force on Process Mining, revealed that PMiner outperforms state-of-the-art methods in detecting business process anomalies. This framework helps businesses to identify process anomalies and rectify them in time to leverage business continuity prospects.

show abstract

“…Bayesian networks are a class of probabilistic graphical models that can be applied to repair an event log with missing timestamps [14] and missing events [15,16]. Additionally, long short-term memory (LSTM) is an artificial neural network in deep learning, able to predict the missing event and activity labels in event logs [17]. Another technology enabling the resolution of missing data issues is likelihood-based algorithms, i.e., single imputation by event relationship (SIER) and multiple imputation by event chain (MIEC), which are able to repair event logs with missing events, timestamps, and resources [18].…”

Section: A Review Of Event Log Preprocessing Techniquesmentioning

confidence: 99%

“…It should be noted that several previously mentioned approaches from the artificial intelligence, machine learning, and deep learning category are able to partly recover missing data. However, these techniques utilize external reference models, i.e., process models defined based on pre-existing process knowledge, and align the incomplete event log according to the expected behavior [15][16][17].…”

Section: A Review Of Event Log Preprocessing Techniquesmentioning

confidence: 99%

Event Log Data Quality Issues and Solutions

et al. 2023

View full text Add to dashboard Cite

Process mining is a discipline that analyzes real event data extracted from information systems that support a business process to construct as-is process models and detect performance issues. Process event data are transformed into event logs, where the level of data quality directly impacts the reliability, validity, and usefulness of the derived process insights. The literature offers a taxonomy of preprocessing techniques and papers reporting on solutions for data quality issues in particular scenarios without exploring the relationship between the data quality issues and solutions. This research aims to discover how process mining researchers and practitioners solve certain data quality issues in practice and investigates the nature of the relationship between data quality issues and preprocessing techniques. Therefore, a study was undertaken among prominent process mining researchers and practitioners, gathering information regarding the perceived importance and frequency of data quality issues and solutions and the participants’ recommendations on preprocessing technique selection. The results reveal the most important and frequent data quality issues and preprocessing techniques and the gap between their perceived frequency and importance. Consequently, an overview of how researchers and practitioners solve data quality issues is presented, allowing the development of recommendations.

show abstract

A Deep Learning Approach for Repairing Missing Activity Labels in Event Logs for Process Mining

Cited by 5 publications

References 42 publications

Log Clustering-based Method for Repairing Missing Traces with Context Probability Information

Log Clustering-based Method for Repairing Missing Traces with Context Probability Information

PMiner: Process Mining using Deep Autoencoder for Anomaly Detection and Reconstruction of Business Processes

Event Log Data Quality Issues and Solutions

Contact Info

Product

Resources

About