駕駛風格被定義為駕駛對複雜駕駛環境的敏感性和車輛控制的掌握程度之間的關係，透過車輛動態分析駕駛風格可作為事故預警和駕駛習慣改進的參考，從而減少道路事故的發生。現行駕駛風格的研究，多以蒐集自然駕駛資料，再進行人工標記特徵和後續分析；然而在真實情境中，大量且即時產生的高維度資料並沒有適合特徵提供後續研究與分析，而且由人工標記特徵不僅成本高、費時且需要相當的領域知識判斷。
為改善現行駕駛風格研究問題，本研究提出一套混合監督與非監督的深度學習框架HyRANet(Hybrid Risk Autoencoder Network)，透過監督式深度時序模型萃取車輛動態特徵作為先驗知識，再由非監督式深度稀疏化自編碼網路將駕駛風格特徵編碼成駕駛風格指跡。經過小量樣本訓練後，便可透過HyRANet編碼器將即時且大量的車輛動態編碼成駕駛風格指跡，編碼後的指跡具有同類相近、不同類疏離的編碼特性，且於低維度資料中保留高維度特徵。在駕駛風格預測上，我們設計以6秒的時間序列預測下6秒的駕駛風格，並比較HyRANet指跡和標準化資料作為輸入的兩種現行機器學習和四種深度學習上的表現。此外，為了處理危險風格偏少所導致的資料不平衡問題，研究中提出使用Focal-loss的改善預測模型，加強少數風格的分類能力。
實驗結果顯示，經t-SNE資料視覺化分布後HyRANet指跡相較於標準化資料更具有可分類性。而使用Focal-loss與HyRANet指跡的模型在六種模型中獲得最佳表現，相較以標準化資料為輸入的隨機森林模型，於難分類的類別提升了約20%的分類效能；另外，使用HyRANet作為為輸入的隨機森林模型，相較以標準化資料為輸入者，於較難分類的類別提升約10%的分類效能，降低約10倍運算時間。
未來可以透過HyRANet風格指跡降低後續資料計算、傳輸與儲存成本，並透過研究中提出的深度預測模型提升分類精準度，應用於風險預測、風格評估等後續分析領域上。

SUMMARY
The driving style is defined as the relationship of the driving sensitivity in complex driving environment and the mastery of vehicle control. The analysis of the driving style can be used as a reference for accident warning and improvement of driving habits, such reducing the occurrence of road Accidents.
However, in real-world situations, large-scale and instantaneously generated high-dimensional materials do not have suitable features to provide subsequent research and analysis, and manual tagging features are not only costly, time-consuming, and require considerable domain knowledge.
In order to improve the current research on driving style, this study proposes a hybrid supervised and unsupervised deep learning framework HyRANet (Hybrid Risk Autoencoder Network), which extracts driving style features in vehicle dynamics as a priori knowledge through supervised deep time series model. Driving style features are then encoded into driving style fingerprint by an unsupervised deep sparse autoencoder network. Based on HyRANet fingerprint, we designed the deep learning framework to predict the driving style with the special loss function.
In the encoding ability experiment, we use t-SNE to visualize the style fingerprint encoded by HyRANet. Compare with the original input data, the style fingerprint have the same kind of similar and heterogeneous separation characteristics. In terms of driving style prediction, the Focal-HyRANet deep learning style prediction model proposed in this paper is about 20% higher in F1-score than the random forest model with standardized data input.

1. Bergasa, L. M., Almería, D., Almazán, J., Yebes, J. J., & Arroyo, R. (2014). Drivesafe: An app for alerting inattentive drivers and scoring driving behaviors. Paper presented at the 2014 IEEE Intelligent Vehicles symposium proceedings.
2. Brombacher, P., Masino, J., Frey, M., & Gauterin, F. (2017). Driving event detection and driving style classification using artificial neural networks. Paper presented at the Industrial Technology (ICIT), 2017 IEEE International Conference on.
3. Carvalho, E., Ferreira, B. V., Ferreira, J., de Souza, C., Carvalho, H. V., Suhara, Y., . . . Pessin, G. (2017). Exploiting the use of recurrent neural networks for driver behavior profiling. Paper presented at the Neural Networks (IJCNN), 2017 International Joint Conference on.
4. Cho, K., Van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078.
5. Dong, W., Li, J., Yao, R., Li, C., Yuan, T., & Wang, L. (2016). Characterizing driving styles with deep learning. arXiv preprint arXiv:1607.03611.
6. Dong, W., Yuan, T., Yang, K., Li, C., & Zhang, S. (2017). Autoencoder regularized network for driving style representation learning. arXiv preprint arXiv:1701.01272.
7. Guo, J., Liu, Y., Zhang, L., & Wang, Y. (2018). Driving Behaviour Style Study with a Hybrid Deep Learning Framework Based on GPS Data. Sustainability, 10(7), 2351.
8. Hallac, D., Sharang, A., Stahlmann, R., Lamprecht, A., Huber, M., Roehder, M., & Leskovec, J. (2016). Driver identification using automobile sensor data from a single turn. Paper presented at the Intelligent Transportation Systems (ITSC), 2016 IEEE 19th International Conference on.
9. Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
10. Klauer, S. G., Dingus, T. A., Neale, V. L., Sudweeks, J. D., & Ramsey, D. J. (2006). The impact of driver inattention on near-crash/crash risk: An analysis using the 100-car naturalistic driving study data.
11. Kluger, R., Smith, B. L., Park, H., & Dailey, D. J. (2016). Identification of safety-critical events using kinematic vehicle data and the discrete fourier transform. Accident Analysis & Prevention, 96, 162-168.
12. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. Paper presented at the Advances in neural information processing systems.
13. LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.
14. Li, G., Li, S. E., Cheng, B., & Green, P. (2017). Estimation of driving style in naturalistic highway traffic using maneuver transition probabilities. Transportation research part C: emerging technologies, 74, 113-125.
15. Li, Y., Xue, F., Feng, L., & Qu, Z. (2017). A driving behavior detection system based on a smartphone's built‐in sensor. International Journal of Communication Systems, 30(8), e3178.
16. Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollár, P. (2017). Focal loss for dense object detection. Paper presented at the Proceedings of the IEEE international conference on computer vision.
17. Liu, H., Taniguchi, T., Takano, T., Tanaka, Y., Takenaka, K., & Bando, T. (2014). Visualization of driving behavior using deep sparse autoencoder. Paper presented at the 2014 IEEE Intelligent Vehicles Symposium Proceedings.
18. Liu, H., Taniguchi, T., Tanaka, Y., Takenaka, K., & Bando, T. (2015). Essential feature extraction of driving behavior using a deep learning method. Paper presented at the 2015 IEEE Intelligent Vehicles Symposium (IV).
19. Liu, H., Taniguchi, T., Tanaka, Y., Takenaka, K., & Bando, T. (2017). Visualization of driving behavior based on hidden feature extraction by using deep learning. IEEE Transactions on Intelligent Transportation Systems, 18(9), 2477-2489.
20. Martinez, C. M., Heucke, M., Wang, F.-Y., Gao, B., & Cao, D. (2017). Driving style recognition for intelligent vehicle control and advanced driver assistance: A survey. IEEE Transactions on Intelligent Transportation Systems.
21. Meseguer, J. E., Toh, C. K., Calafate, C. T., Cano, J. C., & Manzoni, P. (2017). Drivingstyles: a mobile platform for driving styles and fuel consumption characterization. Journal of Communications and Networks, 19(2), 162-168.
22. Misawa, H., Takenaka, K., Sugihara, T., Liu, H., Taniguchi, T., & Bando, T. (2017). Prediction of driving behavior based on sequence to sequence model with parametric bias. Paper presented at the Intelligent Transportation Systems (ITSC), 2017 IEEE 20th International Conference on.
23. Perez, M. A., Sudweeks, J. D., Sears, E., Antin, J., Lee, S., Hankey, J. M., & Dingus, T. A. (2017). Performance of basic kinematic thresholds in the identification of crash and near-crash events within naturalistic driving data. Accident Analysis & Prevention, 103, 10-19.
24. Romera, E., Bergasa, L. M., & Arroyo, R. (2016a). Need data for driver behaviour analysis? Presenting the public UAH-DriveSet. Paper presented at the Intelligent Transportation Systems (ITSC), 2016 IEEE 19th International Conference on.
25. Romera, E., Bergasa, L. M., & Arroyo, R. (2016b). Need data for driver behaviour analysis? Presenting the public UAH-DriveSet. Paper presented at the 2016 IEEE 19th International Conference on Intelligent Transportation Systems (ITSC).
26. Van Ly, M., Martin, S., & Trivedi, M. M. (2013). Driver classification and driving style recognition using inertial sensors. Paper presented at the Intelligent Vehicles Symposium (IV), 2013 IEEE.
27. Wang, J., Zheng, Y., Li, X., Yu, C., Kodaka, K., & Li, K. (2015). Driving risk assessment using near-crash database through data mining of tree-based model. Accident Analysis & Prevention, 84, 54-64.
28. Wijnands, J. S., Thompson, J., Aschwanden, G. D., & Stevenson, M. (2018). Identifying behavioural change among drivers using Long Short-Term Memory recurrent neural networks. Transportation Research Part F: Traffic Psychology and Behaviour, 53, 34-49.
29. Xie, J., Hilal, A. R., & Kulić, D. (2017). Driving maneuver classification: A comparison of feature extraction methods. IEEE Sensors Journal, 18(12), 4777-4784.
30. Xie, J., Hilal, A. R., & Kulić, D. (2018). Driving Maneuver Classification: A Comparison of Feature Extraction Methods. IEEE Sensors Journal, 18(12), 4777-4784.
31. Yosinski, J., Clune, J., Bengio, Y., & Lipson, H. (2014). How transferable are features in deep neural networks? Paper presented at the Advances in neural information processing systems.
32. Zhang, C., Patel, M., Buthpitiya, S., Lyons, K., Harrison, B., & Abowd, G. D. (2016). Driver classification based on driving behaviors. Paper presented at the Proceedings of the 21st International Conference on Intelligent User Interfaces.