Promoting Simple Agents: Ensemble Methods for Event-Log Prediction
For process mining practitioners, this work demonstrates that lightweight models can replace resource-intensive neural networks for event-log prediction, offering a practical trade-off between accuracy and computational cost.
The paper shows that simple n-gram models with appropriate context windows achieve comparable accuracy to LSTM and Transformer models for next-activity prediction in streaming event logs, while requiring substantially fewer resources. The proposed promotion ensemble method dynamically selects between two models during inference, reducing overhead while matching or exceeding neural model accuracy.
We compare lightweight automata-based models (n-grams) with neural architectures (LSTM, Transformer) for next-activity prediction in streaming event logs. Experiments on synthetic patterns and five real-world process mining datasets show that n-grams with appropriate context windows achieve comparable accuracy to neural models while requiring substantially fewer resources. Unlike windowed neural architectures, which show unstable performance patterns, n-grams provide stable and consistent accuracy. While we demonstrate that classical ensemble methods like voting improve n-gram performance, they require running many agents in parallel during inference, increasing memory consumption and latency. We propose an ensemble method, the promotion algorithm, that dynamically selects between two active models during inference, reducing overhead compared to classical voting schemes. On real-world datasets, these ensembles match or exceed the accuracy of non-windowed neural models with lower computational cost.