A trajectory outlier detection method based on variational auto-encoder

Longmei Zhang; Wei Lu; Feng Xue; Yanshuo Chang; Longmei Zhang; Wei Lu; Feng Xue; Yanshuo Chang

doi:10.3934/mbe.2023675

Mathematical Biosciences and Engineering

2023, Volume 20, Issue 8: 15075-15093. doi: 10.3934/mbe.2023675

Previous Article Next Article

Research article Special Issues

A trajectory outlier detection method based on variational auto-encoder

1.
School of Communication and Information Engineering, Xi'an University of Science and Technology, Xi'an 710054, China
2.
School of Information, Xi'an University of Finance and Economics, Xi'an 710100, China

Academic Editor: Yiu-ming Cheung

Received: 15 February 2023 Revised: 27 June 2023 Accepted: 09 July 2023 Published: 14 July 2023

Trajectory outlier detection can identify abnormal phenomena from a large number of trajectory data, which is helpful to discover or predict potential traffic risks. In this work, we proposed a trajectory outlier detection model based on variational auto-encoder. First, the model encodes the trajectory data as parameters of distribution functions based on the statistical characteristics of urban traffic. Then, an auto-encoder network is built and trained. The training goal of the auto-encoder network is to maximize the generation probability of original trajectories when decoding. Once the model training is completed, we can detect the trajectory outlier by the difference between a trajectory and the trajectory generated by the model. The advantage of the proposed model is that it only needs to compute the difference between the original trajectory and the trajectory generated by the model when detecting the trajectory outlier, which greatly reduces the amount of calculation and makes the model very suitable for real-time detection scenarios. In addition, the distance threshold between the abnormal trajectory and the normal trajectory can be set by referring to the proportion of the abnormal trajectory in the training data set, which eliminates the difficulty of setting the threshold manually and makes the model more convenient to be applied in different actual scenes. In terms of effect, the proposed model has achieved more than 95% in accuracy, which is better than the two typical density-based and classification-based detection methods, and also better than the methods based on machine learning in recent years. In terms of efficiency, the model has good convergence in the training phase and the training time increases slowly with the data scale, which is better than or as the same as the comparison methods.
- trajectory outlier detection,
- machine learning,
- variational auto-encoder,
- trajectory similarity
Citation: Longmei Zhang, Wei Lu, Feng Xue, Yanshuo Chang. A trajectory outlier detection method based on variational auto-encoder[J]. Mathematical Biosciences and Engineering, 2023, 20(8): 15075-15093. doi: 10.3934/mbe.2023675

Related Papers:

Abstract

Trajectory outlier detection can identify abnormal phenomena from a large number of trajectory data, which is helpful to discover or predict potential traffic risks. In this work, we proposed a trajectory outlier detection model based on variational auto-encoder. First, the model encodes the trajectory data as parameters of distribution functions based on the statistical characteristics of urban traffic. Then, an auto-encoder network is built and trained. The training goal of the auto-encoder network is to maximize the generation probability of original trajectories when decoding. Once the model training is completed, we can detect the trajectory outlier by the difference between a trajectory and the trajectory generated by the model. The advantage of the proposed model is that it only needs to compute the difference between the original trajectory and the trajectory generated by the model when detecting the trajectory outlier, which greatly reduces the amount of calculation and makes the model very suitable for real-time detection scenarios. In addition, the distance threshold between the abnormal trajectory and the normal trajectory can be set by referring to the proportion of the abnormal trajectory in the training data set, which eliminates the difficulty of setting the threshold manually and makes the model more convenient to be applied in different actual scenes. In terms of effect, the proposed model has achieved more than 95% in accuracy, which is better than the two typical density-based and classification-based detection methods, and also better than the methods based on machine learning in recent years. In terms of efficiency, the model has good convergence in the training phase and the training time increases slowly with the data scale, which is better than or as the same as the comparison methods.

References

[1]	Y. Djenouri, D. Djenouri, J. Lin, Trajectory outlier detection: New problems and solutions for smart cities, ACM Trans. Knowl. Discov. D, 15 (2021), 1–28. https://doi.org/10.1145/3425867 doi: 10.1145/3425867
[2]	Y. Djenouri, A. Belhadi, J. Lin, D. Djenouri, A. Cano, A survey on urban traffic anomalies detection algorithms, IEEE Access, 7 (2019), 12192–12205. https://doi.org/10.1109/ACCESS.2019.2893124 doi: 10.1109/ACCESS.2019.2893124
[3]	U. Ahmed, G. Srivastava, Y. Djenouri, J. C. W. Lin, Deviation point curriculum learning for trajectory outlier detection in cooperative intelligent transport systems, IEEE Trans. Intell. Transp., 23 (2022), 16514–16523. https://doi.org/10.1109/TITS.2021.3131793 doi: 10.1109/TITS.2021.3131793
[4]	F. Meng, G. Yuan, S. Lv, Z. Wang, S. Xia, An overview on trajectory outlier detection, Artif. Intell. Rev., 52 (2019), 2437–2456. https://doi.org/10.1007/s10462-018-9619-1 doi: 10.1007/s10462-018-9619-1
[5]	A. Belhadi, Y. Djenouri, C. Lin, A. Cano, Trajectory outlier detection: Algorithms, taxonomies, evaluation and open challenges, ACM Trans. Manage. Inf., 11 (2020), 1–29. https://doi.org/10.1145/3399631 doi: 10.1145/3399631
[6]	S. Wang, J. Cao, P. Yu, Deep learning for Spatio-Temporal data mining: A survey, IEEE Trans. Knowl. Data Eng., 34 (2022), 3681–3700. https://doi.org/10.1109/TKDE.2020.3025580 doi: 10.1109/TKDE.2020.3025580
[7]	Z. Liu, D. Pi, J. Jiang, Density-based trajectory outlier detection algorithm, J. Syst. Eng. Electron., 24 (2013), 335–340. https://doi.org/10.1109/JSEE.2013.00042 doi: 10.1109/JSEE.2013.00042
[8]	J. Tang, H. Ngan, Traffic outlier detection by density-based bounded local outlier factors, Inf. Technol. Ind., 4 (2016), 6–18. https://doi.org/10.17762/itii.v4i1.38 doi: 10.17762/itii.v4i1.38
[9]	J. Mao, T. Wang, C. Jin, A. Zhou, Feature grouping-based outlier detection upon streaming trajectories, IEEE Trans. Knowl. Data Eng., 29 (2017), 2696–2709. https://doi.org/10.1109/TKDE.2017.2744619 doi: 10.1109/TKDE.2017.2744619
[10]	C. Piciarelli, G. L.Foresti, Anomalous trajectory detection using support vector machines, in IEEE Conference on Advanced Video and Signal Based Surveillance, (2007), 153–158. https://doi.org/10.1109/AVSS.2007.4425302
[11]	P. R. Lei, A framework for anomaly detection in maritime trajectory behavior, Knowl. Inf. Syst., 47 (2016), 189–214. https://doi.org/10.1007/s10115-015-0845-4 doi: 10.1007/s10115-015-0845-4
[12]	J. Wang, Y. Yuan, T. Ni, Y. Ma, M. Liu, G. Xu, et al., Anomalous trajectory detection and classification based on difference and intersection set distance, IEEE Trans. Veh. Technol., 69 (2020), 2487–2500. https://doi.org/10.1109/TVT.2020.2967865 doi: 10.1109/TVT.2020.2967865
[13]	L. X. Pang, S. Chawla, W. Liu, Y. Zheng, On mining anomalous patterns in road traffic streams, Adv. Data Min. Appl., 2011 (2011), 237–251. https://doi.org/10.1007/978-3-642-25856-5_18 doi: 10.1007/978-3-642-25856-5_18
[14]	L. X. Pang, S. Chawla, W. Liu, Y. Zheng, On detection of emerging anomalous traffic patterns using GPS data, Data Knowl. Eng., 97 (2013), 357–373. https://doi.org/10.1016/j.datak.2013.05.002 doi: 10.1016/j.datak.2013.05.002
[15]	Q. Yu, Y. Luo, C. Chen, X. Wang, Trajectory outlier detection approach based on common slices sub-sequence, Appl. Intell., 48 (2018), 2661–2680. https://doi.org/10.1007/s10489-017-1104-z doi: 10.1007/s10489-017-1104-z
[16]	A. Belhadi, Y. Djenouri, D. Djenouri, T. Michalak, J. Lin, Deep learning versus traditional solutions for group trajectory outliers, IEEE Trans. Cybern., 52 (2022), 4508–4519. https://doi.org/10.1109/TCYB.2020.3029338 doi: 10.1109/TCYB.2020.3029338
[17]	Z. Zhang, G. Ni, Y. Xu, Comparison of trajectory clustering methods based on K-means and DBSCAN, in 2020 IEEE International Conference on Information Technology, Big Data and Artificial Intelligence (ICIBA), 1 (2020), 557–561. https://doi.org/10.1109/ICIBA50161.2020.9277214
[18]	W. Dai, C. Zhang, X. Su, S. Cao, Trajectory outlier detection based on DBSCAN and velocity entropy, in iThings and IEEE GreenCom and IEEE CPSCom and IEEE SmartData and IEEE Cybermatics, (2020), 550–557. https://doi.org/10.1109/iThings-GreenCom-CPSCom-SmartData-Cybermatics50389.2020.00097
[19]	C. Lyu, X. Wu, Y. Liu, Z. Liu, A Partial-Fréchet-Distance-Based framework for bus route identification, IEEE Trans. Intell. Transp., 23 (2021), 9275–9280. https://10.1109/TITS.2021.3069630
[20]	B. K. Yi, H. V. Jagadish, C. Faloutsos, Efficient retrieval of similar time sequences under time warping, in IEEE ICDE, (1998), 201−208. https://doi.org/10.1109/ICDE.1998.655778
[21]	M. Vlachos, G. Kollios, D. Gunopulos, Discovering similar multidimensional trajectories, in IEEE ICDE, (2002), 673−684. https://doi.org/10.1109/ICDE.2002.994784
[22]	P. H. Daniel, K. Klara, M. K. Jon, On dynamic Voronoi diagrams and the minimum Hausdorff distance for point sets under Euclidean motion in the plane, in Proceedings of the Eighth Annual Symposium on Computational Geometry, (1992), 110−119. https://doi.org/10.1145/142675.142700
[23]	D. Zhang, Z. Chang, S. Wu, Y. Yuan, K. L. Tan, G. Chen, Continuous trajectory similarity search for online outlier detection, IEEE Trans. Knowl. Data Eng., 34 (2022), 4690–4704. https://doi.org/10.1109/TKDE.2020.3046670 doi: 10.1109/TKDE.2020.3046670
[24]	Y. LeCun, Y. Bengio, G. Hinton, Deep learning, Nature, 521 (2015), 436–444. https://doi.org/10.1038/nature14539 doi: 10.1038/nature14539
[25]	T. Fernando, S. Denman, S. Sridharan, C. Fookes, Soft + Hardwired attention: An LSTM framework for human trajectory prediction and abnormal event detection, Neural Networks, 108 (2018), 466–478. https://doi.org/10.1016/j.neunet.2018.09.002 doi: 10.1016/j.neunet.2018.09.002
[26]	A. Belhadi, Y. Djenouri, G. Srivastava, A. Cano, J. Lin, Hybrid group anomaly detection for sequence data: Application to trajectory data analytics, IEEE Trans. Intell. Transp., 23 (2022), 9346–9357. https://doi.org/10.1109/TITS.2021.3114064 doi: 10.1109/TITS.2021.3114064
[27]	G. E. Hinton, Reducing the dimensionality of data with neural networks, Science, 313 (2006), 504–507. https://doi.org/10.1126/science.1127647 doi: 10.1126/science.1127647
[28]	D. P. Kingma, M. Welling, Auto-encoding variational bayes, preprint, arXiv: 1312.6114.
[29]	X. Chen, J. Xu, R. Zhou, W. Chen, J. Fang, C. Liu, TrajVAE: A Variational AutoEncoder model for trajectory generation, Neurocomputing, 428 (2021), 332–339. https://doi.org/10.1016/j.neucom.2020.03.120 doi: 10.1016/j.neucom.2020.03.120
[30]	J. Chen, S. Sathe, C. Aggarwal, D. Turaga, Outlier detection with autoencoder ensembles, in Proceedings of the 2017 SIAM International Conference on Data Mining, (2017), 90–98. https://doi.org/10.1137/1.9781611974973.11
[31]	D. Hendrycks, M. Mazeika, T. Dietterich, Deep anomaly detection with outlier exposure, preprint, arXiv: 1812.04606.
[32]	P. Malhotra, A. Ramakrishnan, G. Anand, L. Vig, P. Agarwal, G. Shroff, LSTM-based encoder-decoder for multi-sensor anomaly detection, preprint, arXiv: 1607.00148.
[33]	Y. Liu, K. Zhao, G. Cong, Z. Bao, Online anomalous trajectory detection with deep generative sequence modeling, in IEEE ICDE, (2020), 949–960. https://doi.org/10.1109/ICDE48307.2020.00087

Reader Comments

Your name:*

Email:*
© 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)