Dmitrijs Balabka
Denys Shkliarenko
ABSTRACT
Participants of the fourth edition of SHL recognition challenge 2021 aim to recognize eight locomotion and transportation activities in a user-independent manner based on radio data, including GPS reception, GPS location, WiFi reception, and GSM cell tower scans. Team "DD" proposes applying Google's AutoML Tables service to preprocess data, train, and evaluate the model. During this challenge, we showed the advantages and disadvantages of AutoML Tables. In addition, we have employed additional complimentary publically available datasets. AutoML Tables helped to train an artificial neural network model using the AdaNet algorithm. As a result, it has shown an ability to recognize classes with a precision of 81.2% and recall of 78.2%. Also, we opened the source code of the required feature engineering and published it on GitHub: https://github.com/dbalabka/shl-activity-recognition-2021.
- Martín Abadi 2016. TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems. (March 2016). Retrieved from http://arxiv.org/abs/1603.04467Google Scholar
- Sercan O. Arik and Tomas Pfister. 2019. TabNet: Attentive Interpretable Tabular Learning. (2019). Retrieved June 15, 2021 from http://arxiv.org/abs/1908.07442Google Scholar
- Dmitrijs Balabka. 2019. Semi-supervised learning for human activity recognition using adversarial autoencoders. In UbiComp/ISWC 2019- - Adjunct Proceedings of the 2019 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2019 ACM International Symposium on Wearable Computers, 685–688. DOI:https://doi.org/10.1145/3341162.3344854Google ScholarDigital Library
- Dmitrijs Balabka. 2021. SHL challenge 2021 solution source. Retrieved from https://github.com/dbalabka/shl-activity-recognition-2021/Google Scholar
- T. Butler, H., Daly, M., Doyle, A., Gillies, S., Hagen, S., & Schaub. 2016. The GeoJSON Format. Internet Eng. Task Force (2016). Retrieved from https://datatracker.ietf.org/doc/html/rfc7946Google ScholarDigital Library
- Maximilian Christ 2018. Time Series FeatuRe Extraction on basis of Scalable Hypothesis tests (tsfresh – A Python package). Neurocomputing 307, (September 2018), 72–77. DOI:https://doi.org/10.1016/j.neucom.2018.03.067Google ScholarDigital Library
- Department for Transport. 2014. TransXChange. Retrieved from https://www.gov.uk/government/collections/transxchangeGoogle Scholar
- Nick Erickson 2020. AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data. (March 2020). Retrieved from http://arxiv.org/abs/2003.06505Google Scholar
- Matthias Feurer 2015. Efficient and Robust Automated Machine Learning. Adv. Neural Inf. Process. Syst. 28 (2015), 2944–2952. Retrieved June 24, 2018 from http://papers.nips.cc/paper/5872-efficient-and-robust-automated-machine-learning.pdfGoogle Scholar
- Sean Gillies. 2007. Shapely: manipulation and analysis of geometric objects. Toblerity.Org. Retrieved from https://github.com/Toblerity/ShapelyGoogle Scholar
- Hristijan Gjoreski 2018. The University of Sussex-Huawei Locomotion and Transportation Dataset for Multimodal Analytics with Mobile Devices. IEEE Access 6, (2018), 42592–42604. DOI:https://doi.org/10.1109/ACCESS.2018.2858933Google Scholar
- Google. Viewing model architecture with Cloud Logging. Retrieved from https://cloud.google.com/automl-tables/docs/loggingGoogle Scholar
- Google. AI Explanations Whitepaper. Retrieved from https://storage.googleapis.com/cloud-ai-whitepapers/AI Explainability Whitepaper.pdfGoogle Scholar
- Google. 2019. An End-to-End AutoML Solution for Tabular Data at KaggleDays. Retrieved from https://ai.googleblog.com/2019/05/an-end-to-end-automl-solution-for.htmlGoogle Scholar
- Google. 2021. Google Geolocation API. Retrieved from https://developers.google.com/maps/documentation/geolocation/overviewGoogle Scholar
- Google. 2021. Wi-Fi scanning overview. Retrieved from https://developer.android.com/guide/topics/connectivity/wifi-scanGoogle Scholar
- H2O. 2019. A Deep Dive into H2O's AutoML. Retrieved from https://www.h2o.ai/blog/a-deep-dive-into-h2os-automl/Google Scholar
- Xin He 2021. AutoML: A survey of the state-of-the-art. Knowledge-Based Syst. 212, (2021), 106622. DOI:https://doi.org/10.1016/j.knosys.2020.106622Google Scholar
- James Max Kanter and Kalyan Veeramachaneni. 2015. Deep feature synthesis: Towards automating data science endeavors. In 2015 IEEE International Conference on Data Science and Advanced Analytics (DSAA), 1–10. DOI:https://doi.org/10.1109/DSAA.2015.7344858Google ScholarCross Ref
- Anna Kuokkanen 2020. A proposal for a novel urban mobility policy: Personal carbon trade experiment in Lahti city. Util. Policy 62, (February 2020), 100997. DOI:https://doi.org/10.1016/j.jup.2019.100997Google ScholarCross Ref
- Mapbox. 2021. geojson.io. Retrieved from https://geojson.io/Google Scholar
- Mozilla. 2021. Mozilla location service. Retrieved from https://location.services.mozilla.com/Google Scholar
- Randal S. Olson 2016. Automating Biomedical Data Science Through Tree-Based Pipeline Optimization. . 123–137. DOI:https://doi.org/10.1007/978-3-319-31204-0_9Google Scholar
- OpenStreetMap. 2019. OpenStreetMap. Retrieved January 11, 2020 from https://www.openstreetmap.org/Google Scholar
- Chris Thornton 2013. Auto-WEKA: Combined selection and hyperparameter optimization of classification algorithms. Proc. ACM SIGKDD Int. Conf. Knowl. Discov. Data Min. Part F1288, (August 2013), 847–855. DOI:https://doi.org/10.1145/2487575.2487629Google ScholarDigital Library
- University of Sussex. 2021. Sussex-Huawei Locomotion Challenge 2021. Retrieved from http://www.shl-dataset.org/activity-recognition-challenge-2021/Google Scholar
- Lin Wang 2018. Summary of the Sussex-Huawei Locomotion-Transportation Recognition Challenge. In UbiComp/ISWC 2018 - Adjunct Proceedings of the 2018 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2018 ACM International Symposium on Wearable Computers, 1521–1530. DOI:https://doi.org/10.1145/3267305.3267519Google ScholarDigital Library
- Lin Wang 2019. Enabling reproducible research in sensor-based transportation mode recognition with the sussex-huawei dataset. IEEE Access 7, (2019), 10870–10891. DOI:https://doi.org/10.1109/ACCESS.2019.2890793Google Scholar
- Lin Wang 2019. Summary of the Sussex-Huawei locomotion-transportation recognition challenge 2019. UbiComp/ISWC 2019- - Adjun. Proc. 2019 ACM Int. Jt. Conf. Pervasive Ubiquitous Comput. Proc. 2019 ACM Int. Symp. Wearable Comput. (2019), 849–856. DOI:https://doi.org/10.1145/3341162.3344872Google ScholarDigital Library
- Lin Wang 2020. Summary of the sussex-huawei locomotion-transportation recognition challenge 2020. In UbiComp/ISWC 2020 Adjunct - Proceedings of the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2020 ACM International Symposium on Wearable Computers, 351–358. DOI:https://doi.org/10.1145/3410530.3414341Google ScholarDigital Library
- Lin Wang 2021. Locomotion and transportation Mode Recognition from GPS and radio signals: Summary of SHL Challenge 2021. In Proceedings of the 2021 ACM International Joint Conference and 2021 International Symposium on Pervasive and Ubiquitous Computing and Wearable Computers.Google ScholarDigital Library
- Charles Weill 2019. AdaNet: A Scalable and Flexible Framework for Automatically Learning Ensembles. (April 2019). Retrieved from http://arxiv.org/abs/1905.00080Google Scholar
- Brighton&Hove Open Data. Retrieved from https://www.buses.co.uk/open-dataGoogle Scholar
- RadioCells.org. Retrieved from https://radiocells.org/Google Scholar
Recommendations
Human activity recognition using deep transfer learning of cross position sensor based on vertical distribution of data
AbstractSensor-based human activity recognition and health monitoring are attaining great interest in the eye of the researcher as it maintained the privacy of an individual. A model based on the transfer learning for the vertical distribution of cross ...
Human Activity Recognition from Body Sensor Data using Deep Learning
In recent years, human activity recognition from body sensor data or wearable sensor data has become a considerable research attention from academia and health industry. This research can be useful for various e-health applications such as monitoring ...
A survey of video datasets for human action and activity recognition
Highlights Description of datasets for video-based human activity and action recognition. 68 datasets reported: 28 for heterogeneous and 40 for specific human actions. Useful data, such as web for dowloading, published works or ground truth, are ...
Comments