Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Transformer Hawkes Process

About

Modern data acquisition routinely produce massive amounts of event sequence data in various domains, such as social media, healthcare, and financial markets. These data often exhibit complicated short-term and long-term temporal dependencies. However, most of the existing recurrent neural network based point process models fail to capture such dependencies, and yield unreliable prediction performance. To address this issue, we propose a Transformer Hawkes Process (THP) model, which leverages the self-attention mechanism to capture long-term dependencies and meanwhile enjoys computational efficiency. Numerical experiments on various datasets show that THP outperforms existing models in terms of both likelihood and event prediction accuracy by a notable margin. Moreover, THP is quite general and can incorporate additional structural knowledge. We provide a concrete example, where THP achieves improved prediction performance for learning multiple point processes when incorporating their relational information.

Simiao Zuo, Haoming Jiang, Zichong Li, Tuo Zhao, Hongyuan Zha• 2020

Related benchmarks

TaskDatasetResultRank
Event PredictionStackOverflow
ACC45
58
Event Predictiontaxi
RMSEΔt0.37
40
Next event predictionTaobao
Time RMSE0.357
33
Next event predictionAMAZON
RMSE0.612
32
Event PredictionRetweet
RMSE (Time)22.01
28
Next-event time and location predictionEarthquake
Temporal RMSE2.357
27
Long-horizon predictionAMAZON
RMSE (Δt)0.657
26
Event Forecastingtaxi
RMSE0.369
23
Marked Temporal Point ProcessStackOverflow (test)
RMSE0.951
20
Multi-horizon forecastingRetweet
Inter-event Time RMSE18.51
15
Showing 10 of 90 rows
...

Other info

Follow for update