Abstract
Field Computer-implemented systems for detecting employee knowledge deficits by analyzing conformance deviations between employee process execution patterns and expert-derived reference process models, using process mining algorithms on structured workflow event logs. Background Enterprise workflow systems generate detailed event logs recording every step an employee takes in a business process. When an employee repeatedly deviates from the expected process (skipping steps, adding unnecessary steps, executing steps out of order), this pattern can indicate a knowledge deficit in the underlying domain. The approach described here applies industrial process mining techniques to detect these deviations and attribute them to specific knowledge domains. Technical Description The system ingests event logs from enterprise workflow systems (ERP transaction logs, CRM case handling sequences, ticketing system state transitions) via standardized connectors outputting XES-format event streams. Each event record contains: case_id, activity_name, timestamp, employee_id, and optional payload attributes. A reference process model is constructed by filtering event logs to top-quartile performers based on a configurable performance metric (task completion rate, quality score, or a composite). The filtered logs are processed by the Inductive Miner algorithm to discover a sound Petri net process model. The reference model is stored in a process model registry keyed by process type and is rebuilt quarterly. For each employee, the system performs alignment-based conformance checking by computing an optimal alignment between the employee's observed event trace and the reference Petri net using the A-star search algorithm. The alignment produces: a fitness score ranging from 0.0 to 1.0, a list of synchronous moves (matching steps), log moves (steps the employee took that the model does not expect), and model moves (steps the model expects that the employee skipped). Deviation patterns are mapped to knowledge domains via a process-activity-to-skill ontology maintained in the system's skill registry. Each activity in the Petri net is tagged with one or more required knowledge component identifiers. A gap is generated when two conditions are met simultaneously: the deviation count for activities tagged to a specific knowledge domain exceeds a threshold N (default: 5) over a sliding window of W process instances (default: 20), AND the employee's per-domain fitness score is statistically significantly below the reference population mean as determined by Welch's t-test with p less than 0.05. The gap record contains: employee_id, knowledge_domain_id, fitness_score, deviation_count, reference_population_mean, statistical_significance, and contributing_case_ids. Distinguishing Characteristics This system analyzes structured workflow event logs, not natural language communications. The detection mechanism is Petri net alignment conformance checking, not NLP intent classification, entity extraction, or taxonomy mapping. No composite confidence score is computed from linguistic, classifier, recurrence, or tenure signals. The input is XES-format event data, and the analytical engine is the A-star alignment algorithm against a mined Petri net, which is an entirely different computational paradigm.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Davis, Kenneth, "Petri Net Conformance Checking for Knowledge Gap Detection from Enterprise Workflow Event Logs", Technical Disclosure Commons, (March 25, 2026)
https://www.tdcommons.org/dpubs_series/9630