隨著網路科技的進步，物聯網(Internet of Things, IoT)的設備數量也日益增加，逐漸形成萬物聯網之趨勢。在這股趨勢之下，Narrow Band Internet of Things(NB-IoT)便因應而生。NB-IoT是由3rd Generation Partnership Project (3GPP) Release 13所推出的一項具備低成本特性的Low-Power Wide-Area Network(LPWAN)傳輸技術。在NB-IoT的應用下，智慧農業、智慧工業、智慧城市及智慧車載等物聯網產業也得以蓬勃發展。然而，萬物聯網便代表著龐大的感測資訊量，如何有效率地使用這些數據，也就是「大數據(Big Data)」的應用，便成了一項重要的議題。因此人工智慧的分析變成了一道備受矚目的關鍵技術。在這些應用議題中，駕駛行為分析便是一項利用了物聯網及大數據分析的智慧應用服務，藉由裝載著大量感測器的車上單元(On-Board Unit, OBU)長時間感測駕駛者的駕駛資訊，例如速度、加速度等，並且透過人工智慧的演算法來加以分析，藉此計算出駕駛者的駕駛安全程度，並據此給予相對應的保險折扣。
本研究開發了NB-IoT智慧車載系統，並使用Random Forest提出了一駕駛事件分類機制。本研究中的NB-IoT智慧車載系統，是一個車聯網的無線感測網路系統應用，其主體便是一專門用於感測機車駕駛資訊的OBU。在這OBU中裝載了加速度計、陀螺儀及全球定位系統(Global Positioning System, GPS)模組等感測器。機車的相關感測資訊可以透過此OBU上的NB-IoT網路模組上傳至雲端，以供資料分析之用。在資料收集齊全後透過本研究的事件偵測模組，可以從駕駛資訊中擷取出數段的駕駛事件。為了解決資訊中的資料重疊(Overlapping)以及數據不平衡的問題，在本研究中使用了AutoEncoder-Decoder以及Synthetic Minority Over-sampling Technique (SMOTE)來預處理駕駛事件資訊。接著，以Random Forest加以分析，評估出該駕駛事件是安全或是危險。最後，透過本研究所提出的分析模型便可計算出駕駛的安全程度，並提供給保險公司，藉此實現駕駛行為分析的應用。
本研究以兩種駕駛事件分類模型來進行數據分析與探討，分別為橫向事件分類模型(Lateral Events Classification Model) 與縱向事件分類模型(Longitudinal Events Classification Model)。在駕駛事件分類模型的分析中，橫向事件的分析準確率達到93.28%，標準差為0.9%；縱向事件的分析準確率達到94.03%，標準差為2%。此外，亦針對其他駕駛行為分析相關研究進行比較，在「ANN Driving Style Classification」的方法中所使用的分類演算法為Artificial Neural Network(ANN)，橫向事件的分析準確率為91%；縱向事件的分析準確率為92%。由此實驗結果得知，本研究所提出之機器學習駕駛事件分類機制中，透過使用AutoEncoder-Decoder、SMOTE演算法以及Random Forest演算法確實能解決過度擬合、數據不平衡及分析時間太長等問題，以此達成了較高的準確率。

Due to the advancement of the Internet technology, the amount of Internet of Things(IoT) devices is increasing and the tendency of IoT is also gradually forming. In this case, Narrow Band Internet of Things (NB-IoT) has appeared. NB-IoT is a Low-Power, Wide-Area Network(LPWAN) communication technology with low cost introduced by 3rd Generation Partnership Project (3GPP) Release 13. With the applications of NB-IoT, the industries of IoT such as intelligence agriculture, smart industry, smart city and the Internet of Vehicle (IoV), have also flourished. However, the Internet of Every Things represents a huge amount of sensing information. How to effectively use this data, that is, the application of "Big Data", has become an important issue. Therefore, the analysis of artificial intelligence has become a highly popular technology. Among these application issues, driving style analysis is an intelligent service that utilizes IoT devices and big data. The driving information sensed by the On Board Unit (OBU) can be used to calculate the driving safety degree of the driver and provide insurance discount accordingly.
This study developed the NB-IoT telematics system and proposed the random forest driving event classification mechanism. The NB-IoT telematics system in this study is an OBU applied in IoV. The sensors in the OBU include accelerometer, gyroscope and Global Positioning System (GPS). The motorcycle sensor data can be uploaded to cloud through NB-IoT network module for data analyzing. After data is collected completely, it can extract driving events section by section from the motorcycle sensor data. In order to solve the problem of overlapping and class-imbalance, AutoEncoder-Decoder and Synthetic Minority Over-sampling Technique (SMOTE) is used to preprocess the driving events data. Afterwards, this study uses Random Forest to analyze the driving data to classify the driving event as defensive or sporty. Finally, the safety degree can be calculated by the models. The result of the classification can be provided for insurance companies to achieve the application service of driving style analysis.
This study analyzes driving data by two driving classification models include lateral events classification model and longitudinal events classification model. In the analysis of driving event classification model, the analysis accuracy of lateral events classification model is 93.28% and the standard deviation is 0.9%; the analysis accuracy of longitudinal events classification model is 94.03% and the standard deviation is 2%. In addition, this study also compares to other research of driving behavior. In “ANN Driving Style Classification,” which used Artificial Neural Network (ANN) as the classification algorithm. The analysis accuracy of lateral events classification model is 91% and the analysis accuracy of longitudinal events classification model is 92%. The experimental result shows that the problem of overlapping, class-imbalance and too long training time can really be solved in this study thereby achieve the higher accuracy.

[1] Z. Xiao, X. Fu, L. Zhang, L. Ponnambalam and R. S. M. Goh, “Data-driven Multi-Agent System for Maritime Traffic Safety Management,” Proceedings of the IEEE 20th International Conference on Intelligent Transportation Systems (ITSC), pp. 1-6, 2017.
[2] M. H. A. Awadalla, “Design of a Smart Traffic Information System,” Proceedings of the International Conference on Intelligent Computing, Instrumentation and Control Technologies (ICICICT), pp. 757-762, 2017.
[3] L. Zhou, Q. Chen, Z. Luo, H. Zhu and C. Chen, “Speed-based Location Tracking in Usage-Based Automotive Insurance,” Proceedings of the IEEE 37th International Conference on Distributed Computing Systems (ICDCS), pp. 2252-2257, 2017.
[4] J. Peng, N. Liu, H. Zhao and M. Yu, “Usage-based Insurance System Based on Carrier-cloud-client,” Proceedings of the 10th International Conference on Communications and Networking in China (ChinaCom), pp. 579-584, 2015.
[5] I. Skog, P. Händel, M. Ohlsson and J. Ohlsson, “Challenges in Smartphone-driven Usage Based Insurance,” Proceedings of the IEEE Global Conference on Signal and Information Processing, Austin, pp. 1135-1135, 2013.
[6] P. Händel, J. Ohlsson, M. Ohlsson, I. Skog and E. Nygren, “Smartphone-based Measurement Systems for Road Vehicle Traffic Monitoring and Usage Based Insurance,” IEEE Systems Journal, Vol. 8, No. 4, pp. 1238-1248, December 2014.
[7] J. P. Bardyn, T. Melly, O. Seller and N. Sornin, “IoT: The Era of LPWAN Is Starting Now,” Proceedings of the European Solid-State Circuits Conference, pp. 25-30, 2016.
[8] 3GPP TS 36.321 V13.1.0, June 2016; Introduction of NB-IoT in 36.321.
[9] M. Lauridsen, H. Nguyen, B. Vejlgaard, I. Z. Kovacs, P. Mogensen and M. Sorensen, “Coverage Comparison of GPRS, NB-IoT, LoRa, and SigFox in a 7800 km² Area,” Proceedings of the IEEE 85th Vehicular Technology Conference (VTC Spring), pp. 1-5, 2017.
[10] S. Persia, C. Carciofi and M. Faccioli, “NB-IoT and LoRA Connectivity Analysis for M2M/IoT Smart Grids Applications,” Proceedings of the International Annual Conference, pp. 1-6, 2017.
[11] B. Vejlgaard, M. Lauridsen, H. Nguyen, I. Z. Kovacs, P. Mogensen and M. Sorensen, “Coverage and Capacity Analysis of Sigfox, LoRa, GPRS, and NB-IoT,” Proceedings of the IEEE 85th Vehicular Technology Conference (VTC Spring), pp. 1-5, 2017.
[12] N. S. Knyazev, V. A. Chechetkin and D. A. Letavin, “Comparative Analysis of Standards for Low-power Wide-area Network,” Proceedings of the Systems of Signal Synchronization, Generating and Processing in Telecommunications (SINKHROINFO), pp. 1-4, 2017.
[13] A. Adhikary, X. Lin and Y. P. E. Wang, “Performance Evaluation of NB-IoT Coverage,” Proceedings of the IEEE 84th Vehicular Technology Conference (VTC-Fall), pp. 1-5, 2016.
[14] M. Lauridsen, I. Z. Kovacs, P. Mogensen, M. Sorensen and S. Holst, “Coverage and Capacity Analysis of LTE-M and NB-IoT in a Rural Area,” Proceedings of the IEEE 84th Vehicular Technology Conference (VTC-Fall, pp. 1-5, 2016.
[15] H. Kim, S. Cho, J. Oh and G. Jo, “Analysis of LTE and NB-IoT coexistence,” Proceedings of the International Conference on Information and Communication Technology Convergence (ICTC), pp. 870-872, 2017.
[16] A. Hoglund et al., “Overview of 3GPP Release 14 Enhanced NB-IoT,” IEEE Network, Vol. 31, No. 6, pp. 16-22, November 2017.
[17] F. J. Grión, G. O. Petracca, D. F. Lipuma and E. R. Amigó, “LoRa Network Coverage Evaluation in Urban and Densely Urban Environment Simulation and Validation Tests in Autonomous City of Buenos Aires,” Proceedings of the XVII Workshop on Information Processing and Control (RPIC), pp. 1-5, 2017.
[18] N. Poursafar, M. E. E. Alahi and S. Mukhopadhyay, “Long-range Wireless Technologies for IoT Applications: A Review,” Proceedings of the Eleventh International Conference on Sensing Technology (ICST), pp. 1-6, 2017.
[19] P. Brombacher, J. Masino, M. Frey and F. Gauterin, “Driving Event Detection and Driving Style Classification Using Artificial Neural Networks,” Proceedings of the IEEE International Conference on Industrial Technology (ICIT), pp. 997-1002, 2017.
[20] H. Hejazi, H. Rajab, T. Cinkler and L. Lengyel, “Survey of Platforms for Massive IoT,” Proceedings of the IEEE International Conference on Future IoT Technologies (Future IoT), pp. 1-8, 2018.
[21] F. Bing, “The Research of IOT of Agriculture Based on Three Layers Architecture,” Proceedings of the 2nd International Conference on Cloud Computing and Internet of Things (CCIOT), pp. 162-165, 2016.
[22] C. L. Zhong, Z. Zhu and R. G. Huang, “Study on the IOT Architecture and Gateway Technology,” Proceedings of the 14th International Symposium on Distributed Computing and Applications for Business Engineering and Science (DCABES), pp. 196-199, 2015.
[23] S. E. Khateeb, “IoT Architecture a Gateway for Smart Cities in Arab World,” Proceedings of the 15th Learning and Technology Conference (L&T), pp. 153-160, 2018.
[24] 3GPP TS 36.101 V13.4.0, April 2017; User Equipment (UE) Radio Transmission and Reception.
[25] 3GPP TS 24.008 V15.2.0, September 2016; Mobile Radio Interface Layer 3 Specification.
[26] K. Zfnebi, N. Souissi and K. Tikito, “Driver Behavior Quantitative Models: Identification and Classification of Variables,” Proceedings of the International Symposium on Networks, Computers and Communications (ISNCC), pp. 1-6, 2017.
[27] H. Shirazi, K. Haefner and I. Ray, "Fresh-Phish: A Framework for Auto-Detection of Phishing Websites," Proceedings of the IEEE International Conference on Information Reuse and Integration (IRI), pp. 137-143, 2017
[28] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE: Synthetic Minority Over-sampling Technique,” Journal of Artificial Intelligence, Vol. 16, No. 1, pp. 321-357, June 2002.
[29] A. Paul, D. P. Mukherjee, P. Das, A. Gangopadhyay, A. R. Chintha and S. Kundu, “Improved Random Forest for Classification,” IEEE Transactions on Image Processing, Vol. 27, No. 8, pp. 4012-4024, August 2018.
[30] D. A. Johnson and M. M. Trivedi, “Driving Style Recognition Using a Smartphone as a Sensor Platform,” Proceedings of the 14th International IEEE Conference on Intelligent Transportation Systems (ITSC), pp. 1609-1615, 2011.