駕駛行為車險服務(Usage-based insurance, UBI)是一種新興的汽車保險服務，基於個人駕駛行為及駕駛風險的不同，而反應在客製化的保險費率上。從整體社會的角度來看，UBI可以有效降低用路人的交通風險，提升整體道路的安全水準；從保險業者的觀點而言，推出UBI車險服務能比同業更具有競爭力。然而現行的UBI文獻中仍沒有以機車為標的相關研究，且台灣的特殊地域與人文環境，造就了不同於國外機車的駕駛特性；此外，隱私權、個資安全意識的提升也造成UBI的推行困難。若是推出以機車為標的的UBI車險服務，把UBI車險服務的優勢複製於機車車險市場上，將能創造更好的行車環境，緩解機車傷亡事故產生，亦可以使高事故率的機車族群擁有更好的安全保障。
為突破機車UBI研究匱乏的困境，本研究提出一機車車險服務架構，其中包含提出適合機車的多維度駕駛風險等級、建立駕駛風險指跡模型(Risk-aware fingerprint model)。本研究以三軸加速度、三軸角速度探討機車的操作行為，提出一屬於機車標的的風險等級；利用神經網路之運算特性，透過自編碼神經網路(Autoencoder)萃取車輛動態資訊，將高維原始數據壓縮編碼成低維度的駕駛風險指跡，並透過T-SNE視覺化檢視資料特徵萃取效果，再透過不同深度學習與機器學習進行駕駛風險等級之預測，找出最佳駕駛風險指跡模型。本研究提出的模型及想法，不僅可有效改善過往研究的限制，亦可幫助突破過往UBI推行之困難，從根本改善駕駛人行為，並且提升道路安全。

Usage Based Insurance (UBI) is a trend in auto insurance industry over the world which is based on individual driving behavior and driving risk, reflecting in the customized insurance rates. From the perspective of society, UBI service can effectively reduce the risks of passers-by and improve the overall road safety level. From the perspective of the insurance industry, the introduction of UBI service can be more competitive than competitions. However, there is still no relevant research on motorcycles in the current UBI literature, and Taiwan’s geographical and cultural environment has created special driving characteristics that are different from those of motorcycles in foreign; in addition, the increase in privacy and personal safety awareness has also caused UBI’s difficulty in implementation. If the UBI service with motorcycles as the target is launched, the advantages of UBI service in the automobile insurance market will be copied to the motorcycle insurance market. Also, it will create a better driving environment, alleviate motorcycle casualties, and also enable the drivers protected by insurance.
In order to break through the lack of research on motorcycle UBI, this research proposes a UBI service framework for motorcycles, including proposing multi-dimensional driving risk level suitable for motorcycles and proposing risk-aware fingerprint model. This research explores the operation of the motorcycle with three-axis acceleration and three-axis angular velocity and proposes a risk level standard belonging to the motorcycle; Moreover, using the computational characteristics of the neural network, the dynamic driving information is extracted through the autoencoder. The original data with high-dimensional information is compressed and encoded into low-dimensional driving risk-aware fingerprint (RAF). The performance of feature extraction by autoencoder is visually inspected through T-SNE. Additionally, using RAF with different deep learning and machine learning classifiers to predict risk level and find the best driving risk-aware fingerprint model. The models and ideas proposed in this study can not only effectively improve the limitations of previous studies, but also help break through the difficulties of the implementation of UBI in the past, fundamentally improve driver behavior, and improve road safety.

Attal, F., et al. (2014). Powered two-wheeler riding pattern recognition using a machine-learning framework. IEEE Transactions on Intelligent Transportation Systems, 16(1), 475-487.
Attal, F., et al. (2018). Powered two-wheelers critical events detection and recognition using data-driven approaches. IEEE Transactions on Intelligent Transportation Systems, 19(12), 4011-4022.
Baecke, P., et al. (2017). The value of vehicle telematics data in insurance risk selection processes. Decision Support Systems, 98, 69-79.
Bian, Y., et al. (2018). Good drivers pay less: A study of usage-based vehicle insurance models. Transportation research part A: policy and practice, 107, 20-34.
Chen, C., et al. (2019). A graphical modeling method for individual driving behavior and its application in driving safety analysis using GPS data. Transportation research part F: traffic psychology and behaviour, 63, 118-134.
Chen, J., et al. (2019). Driving safety risk prediction using cost-sensitive with nonnegativity-constrained autoencoders based on imbalanced naturalistic driving data. IEEE Transactions on Intelligent Transportation Systems, 20(12), 4450-4465.
Dong, W., et al. (2016). Characterizing driving styles with deep learning. arXiv preprint arXiv:1607.03611.
Dong, W., et al. (2017). Autoencoder regularized network for driving style representation learning. arXiv preprint arXiv:1701.01272.
Hsiao, C. L. (2019). Design of a Usage-Based Insurance Platform for Evaluating Driver’s Risk by Big Data Analysis (碩士), 國立成功大學, 台南市. Retrieved from https://hdl.handle.net/11296/527gz7
Hu, J., et al. (2020). Abnormal driving detection with normalized driving behavior data: a deep learning approach. IEEE Transactions on Vehicular Technology.
Huang, Y., et al. (2019). Automobile insurance classification ratemaking based on telematics driving data. Decision Support Systems, 127, 113156.
Kurylowicz, L. (2016). Usage-Based Insurance: the concept and study of available analyses. Wiadomości Ubezpieczeniowe, 4.
Lin, J., et al. (2007). Experiencing SAX: a novel symbolic representation of time series. Data Mining and knowledge discovery, 15(2), 107-144.
Ma, Y.-L., et al. (2018). The use of context-sensitive insurance telematics data in auto insurance rate making. Transportation research part A: policy and practice, 113, 243-258.
Mantouka, E. G., et al. (2019). Identifying driving safety profiles from smartphone data using unsupervised learning. Safety Science, 119, 84-90.
Nuswantoro, F. M., et al. (2020). Abnormal Driving Detection Based on Accelerometer and Gyroscope Sensor on Smartphone using Artificial Neural Network (ANN) Algorithm. Paper presented at the 2020 International Electronics Symposium (IES).
Paefgen, J., et al. (2014). Multivariate exposure modeling of accident risk: Insights from Pay-as-you-drive insurance data. Transportation research part A: policy and practice, 61, 27-40.
Sasidhar, K., et al. (2019). Two wheeler rash drive detection using smartphones. Paper presented at the 2019 11th International Conference on Communication Systems & Networks (COMSNETS).
Wu, S.-s. (2020). Identify Highway Crash Hotspot Based on Driving Behavior Data Analyzing.
Zhang, J., et al. (2019). Attention-Based Convolutional and Recurrent Neural Networks for Driving Behavior Recognition Using Smartphone Sensor Data. IEEE Access, 7, 148031-148046.
Zhou, L., et al. (2018). Location privacy in usage-based automotive insurance: Attacks and countermeasures. IEEE Transactions on Information Forensics and Security, 14(1), 196-211.