Graph Grounded Cross Attention Transformer Neural Network for Structurally Constrained Full Event Sequence Generation in Predictive Process Monitoring

2026-06-17

AI Summary

This work addresses the challenge of generating complete event sequences under structural constraints in predictive process monitoring by proposing GGATN, a novel model that jointly generates activities, timestamps, sequence lengths, and multi-level attributes while ensuring compliance with transition feasibility, temporal ordering, termination conditions, and attribute consistency. GGATN integrates a global process graph as structured memory and combines Transformer self-attention with graph-guided cross-attention to inject process topology into generation. Adopting a non-autoregressive paradigm, it employs Viterbi-style graph-constrained decoding to eliminate hallucinations. Evaluated on six benchmark event logs, GGATN substantially outperforms large language model baselines, achieving superior performance in sequence similarity, control-flow fidelity, and temporal distribution accuracy, with zero hallucinated activities and zero sequence-level attribute inconsistencies.

Abstract

Structurally constrained event sequence generation remains challenging because generated paths must preserve transition feasibility, temporal order, termination, and attribute consistency. In predictive process monitoring (PPM), this challenge appears as full event sequence generation, whereas existing work mainly addresses component tasks such as next activity, remaining time, outcome, and attribute prediction. This paper proposes the Graph Grounded Cross Attention Transformer Neural Network (GGATN) for this unified PPM task. GGATN uses a global process graph as structured activity memory, contextualizes sequence positions through Transformer self attention, and injects process topology through graph grounded cross attention. Unlike autoregressive decoding, GGATN generates activities, timestamps, length, and event level and sequence level attributes in a single pass, followed by Viterbi style graph constrained decoding for feasible paths and explicit termination. Experiments on six benchmark event logs show more reliable generation quality than local instruction prompted LLM baselines. GGATN achieves strong performance on sequence similarity, Damerau Levenshtein similarity, bigram based control flow similarity, and duration distribution, while maintaining zero hallucinated activities and zero sequence level attribute inconsistency. Ablation analyses confirm the global graph encoder as a stable structural prior. Interpretability analyses show how graph structure, sequence context, feedback refinement, and constrained decoding shape generation.

Problem

• predictive process monitoring
• structurally constrained generation
• event sequence generation
• process graph
• full sequence prediction

Innovation

• Graph Grounded Cross Attention
• Full Event Sequence Generation
• Predictive Process Monitoring
• Non-autoregressive Generation
• Graph-constrained Decoding