Abstract
Event sequences play as a general fine-grained representation for temporal asynchronous event streams. However, in practice, event sequences are often fragmentary and incomplete with censored intervals or missing data, making it hard for downstream prediction and decision-making tasks. In this work, we propose a fresh extension on the definition of the temporal point process, which conventionally characterizes chronological prediction based on historical events, and introduce inverse point process that characterizes counter-chronological attribution based on future events. These two point process models allow us to impute missing events for one partially observed sequence with conditional intensities in two symmetric directions. We further design a peer imitation learning algorithm that lets two models cooperatively learn from each other, leveraging imputed sequences given by the counterpart as the supervised signal. The training process consists of iterative learning of two models and facilitates them to achieve a consensus. We conduct extensive experiments on both synthetic and real-world datasets, which demonstrate that our model can recover incomplete event sequences very close to the ground-truth, with averagely 49.40% improvement compared with related competitors measured by normalized optimal transport distance.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Antoniou, A.: Digital Signal Processing. McGraw-Hill, New York (2016)
Crites, R.H., Barto, A.G.: Improving elevator performance using reinforcement learning. In: NeurIPS, pp. 1017–1023 (1995)
Daley, D.J., Vere-Jones, D.: An Introduction to the Theory of Point Processes: Volume II: General Theory and Structure. Springer, Heidelberg (2007). https://doi.org/10.1007/978-0-387-49835-5
Du, N., Dai, H., Trivedi, R., Upadhyay, U., Gomez-Rodriguez, M., Song, L.: Recurrent marked temporal point processes: embedding event history to vector. In: SIGKDD, pp. 1555–1564 (2016)
Enguehard, J., Busbridge, D., Bozson, A., Woodcock, C., Hammerla, N.Y.: Neural temporal point processes for modelling electronic health records (2020)
Fan, Y., Xu, J., Shelton, C.R.: Importance sampling for continuous time bayesian networks. J. Mach. Learn. Res. 11(Aug), 2115–2140 (2010)
Goodfellow, I.J., et al.: Generative adversarial nets. In: NeurIPS, pp. 2672–2680 (2014)
Hawkes, A.G.: Spectra of some self-exciting and mutually exciting point processes. Biometrika 58(1), 83–90 (1971)
Ho, J., Ermon, S.: Generative adversarial imitation learning. In: NeurIPS, pp. 4565–4573 (2016)
Isham, V., Westcott, M.: A self-correcting point process. Stochastic Process. Appl. 8(3), 335–347 (1979)
Kullback, S., Leibler, R.A.: On information and sufficiency. Ann. Math. Stat. 22(1), 79–86 (1951)
Lee, Y., Vo, T.V., Lim, K.W., Soh, H.: Z-transforms and its inference on partially observable point processes. In: IJCAI, pp. 2369–2375 (2018)
Li, S., Xiao, S., Zhu, S., Du, N., Xie, Y., Song, L.: Learning temporal point processes via reinforcement learning. In: NeurIPS, pp. 10781–10791 (2018)
Mei, H., Qin, G., Eisner, J.: Imputing missing events in continuous-time event streams. In: ICML, pp. 4475–4485 (2019)
Nodelman, U., Shelton, C.R., Koller, D.: Continuous time bayesian networks. arXiv preprint arXiv:1301.0591 (2012)
Pan, Z., Huang, Z., Lian, D., Chen, E.: A variational point process model for social event sequences. In: AAAI, pp. 173–180 (2020)
Rao, V., Teh, Y.W.: MCMC for continuous-time discrete-state systems. In: NeurIPS, pp. 701–709 (2012)
Reinhart, A.: A review of self-exciting spatio-temporal point processes and their applications. Stat. Sci. 33(3), 299–318 (2018)
Schaubel, D.E., Cai, J.: Multiple imputation methods for recurrent event data with missing event category. Can. J. Stat. 34(4), 677–692 (2006)
Shelton, C.R., Qin, Z., Shetty, C.: Hawkes process inference with missing data. In: AAAI, pp. 6425–6432 (2015)
Stomakhin, A., Short, M.B., Bertozzi, A.L.: Reconstruction of missing data in social networks based on temporal patterns of interactions. Inverse Prob. 27(11), 115013 (2011)
Upadhyay, U., De, A., Rodriguez, M.G.: Deep reinforcement learning of marked temporal point processes. In: NeurIPS, pp. 3172–3182 (2018)
Whong, C.: Foiling nyc’s taxi trip data (2014)
Wu, Q., Zhang, Z., Gao, X., Yan, J., Chen, G.: Learning latent process from high-dimensional event sequences via efficient sampling. In: NeurIPS, pp. 3842–3851 (2019)
Xiao, S., Yan, J., Yang, X., Zha, H., Chu, S.M.: Modeling the intensity function of point process via recurrent neural networks. In: AAAI, pp. 1597–1603 (2017)
Zhao, Y., Jiang, H., Wang, X.: Minimum edit distance-based text matching algorithm. In: NLPKE, pp. 1–4 (2010)
Zhou, K., Zha, H., Song, L.: Learning triggering kernels for multi-dimensional hawkes processes. In: ICML, pp. 1301–1309 (2013)
Acknowledgement
This work was supported by the National Key R&D Program of China [2020YFB1707900]; the National Natural Science Foundation of China [62272302, 62172276], and Shanghai Municipal Science and Technology Major Project [2021SHZDZX0102].
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Chen, J., Ye, H., Gao, X., Wu, F., Kong, L., Chen, G. (2023). An Adaptive Data-Driven Imputation Model for Incomplete Event Series. In: Yang, X., et al. Advanced Data Mining and Applications. ADMA 2023. Lecture Notes in Computer Science(), vol 14176. Springer, Cham. https://doi.org/10.1007/978-3-031-46661-8_1
Download citation
DOI: https://doi.org/10.1007/978-3-031-46661-8_1
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-46660-1
Online ISBN: 978-3-031-46661-8
eBook Packages: Computer ScienceComputer Science (R0)