Graph Grounded Cross Attention Transformer Neural Network for Structurally Constrained Full Event Sequence Generation in Predictive Process Monitoring
Researchers have developed a new neural network architecture called the Graph Grounded Cross Attention Transformer Neural Network (GGATN) for predictive process monitoring. This model aims to generate complete event sequences while ensuring structural feasibility, temporal order, and attribute consistency. Unlike previous methods that focused on individual prediction tasks, GGATN generates activities, timestamps, and attributes in a single pass, followed by a constrained decoding step to ensure valid paths. Experiments on benchmark event logs demonstrate GGATN's superior performance in sequence similarity and control flow accuracy compared to LLM baselines, with no hallucinated activities or attribute inconsistencies. AI
IMPACT This new model could improve the accuracy and reliability of predictive process monitoring systems by generating more realistic event sequences.