Learning Hawkes Processes from Short Doubly-Censored Event Sequences
This work addresses a critical bottleneck in quantitative asynchronous event sequence analysis for applications with incomplete data, representing an incremental improvement through a novel data synthesis approach.
The paper tackles the problem of learning Hawkes processes from short doubly-censored event sequences, a common incomplete data scenario, by proposing a sampling-stitching data synthesis method that improves learning results for both time-invariant and time-varying models, as demonstrated in experiments on synthetic and real-world data.
Many real-world applications require robust algorithms to learn point processes based on a type of incomplete data --- the so-called short doubly-censored (SDC) event sequences. We study this critical problem of quantitative asynchronous event sequence analysis under the framework of Hawkes processes by leveraging the idea of data synthesis. Given SDC event sequences observed in a variety of time intervals, we propose a sampling-stitching data synthesis method --- sampling predecessors and successors for each SDC event sequence from potential candidates and stitching them together to synthesize long training sequences. The rationality and the feasibility of our method are discussed in terms of arguments based on likelihood. Experiments on both synthetic and real-world data demonstrate that the proposed data synthesis method improves learning results indeed for both time-invariant and time-varying Hawkes processes.