成功大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	蕭至良 Hsiao, Chih-Liang
論文名稱：	基於大數據分析之駕駛風險評估與駕駛行為車險服務平台設計 Design of a Usage-Based Insurance Platform for Evaluating Driver’s Risk by Big Data Analysis
指導教授：	李威勳 Lee, Wei-Hsun
學位類別：	碩士 Master
系所名稱：	管理學院 - 電信管理研究所 Institute of Telecommunications Management
論文出版年：	2019
畢業學年度：	107
語文別：	英文
論文頁數：	101
中文關鍵詞：	駕駛行為車險應用服務、駕駛風險分析、大數據分析、電信加值服務、瀕臨撞擊事件
外文關鍵詞：	Usage-Based Insurance, Driver risk Assessment, Big Data Analysis, Telecom Value-Added Service, Near Crash Incident
相關次數：	點閱：183 下載：10
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

駕駛行為車險應用服務 (Usage-based insurance)是在全球車輛保險產業中一個顯著的趨勢。其核心概念是基於駕駛人真實的駕駛狀況來預測其駕駛風險並給予駕駛人相對應的客製化保費。先前研究指出UBI可以有效地消除車險市場中的資訊不對稱並且使車險公司、保戶和整體社會都受益。然而，由於目前車險公司普遍受困於資訊不對稱所帶來的惡性循環，導致缺少資金和資源來建立一套完整的UBI系統。因此，本研究提出一個經由電信公司來營運的創新UBI平台來解決車險公司目前所遇到的困境。近年由於OTT(Over The Top)服務的興起，數位轉型是目前電信業普遍的趨勢，加上其本業在3G/4G通訊上擁有成本優勢和租借設備的經驗，本研究認為電信公司可以很好地扮演經營UBI平台的關鍵腳色。另一方面，由於現行的UBI 模型在預測駕駛風險上擁有許多限制與缺點，本研究提出了一個結合資料探勘和瀕臨撞擊事件(Near crash)的新型駕駛特徵：「Driving Pattern-N」來預測駕駛人的行車風險。本研究設計了三個實驗和兩種駕駛風險等級來評估「Driving Pattern-N」、「Driving Pattern」、「Behavior-Centric」和「歷史紀錄」等駕駛特徵在不同情況下對於駕駛風險等級預測的表現。

Usage-based insurance (UBI) is a noticeable trend in auto insurance industry over the world. The core concept of UBI is offering driver customized premium which based on their predicted driving risk according to realistic driving status. Previous research indicated that UBI can effectively eliminate the information asymmetry in auto insurance market and bring benefits for auto insurers, customers and the whole society. However, auto insurers are now lack of resources and funds to establish a complete UBI system since they are suffering from a vicious cycle which is caused by the information asymmetry. Therefore, this research proposed a novel UBI platform and selected a telecommunication company as operator to solve this problem. Recently, owing to the rise of OTT (Over-The-Top) services, digital transformation is a common trend in telecom industry. Moreover, with the cost advantage of data communication and possessing experiences of renting devices, this study believes that telecom company can play an essential role in manipulating the UBI platform. On the other hand, due to the limitations and disadvantages of driving risk prediction on current UBI models, this work purposed a new driving feature－“Driving Pattern-N” which combined data mining techniques and near crash incidents to predict driver risk. Three experiments and two criteria of risk level were designed to evaluate the performance of different driving features (i.e. “Driving pattern-N”, “Driving Pattern”, “Behavior-Centric” and “Historical Record”) under different situations.

Table of Contents
Abstract ii
Acknowledgements iii
Table of Contents iv
List of Tables vii
List of Figures	ix
Chapter 1 Introduction 1
1.1 Research Background 1
1.2 Motivations 8
1.2.1 Limitations of Pay-As-You-Drive model 8
1.2.2 Limitations of Pay-How-You-Drive model 8
1.2.3 Limitations of auto insurers to operate UBI program 9
1.2.4 Design a better PHYD model by driving pattern mining 9
1.2.5 UBI platform operate by a telecom company 10
1.3 Goals 11
1.4 Research Framework 12
Chapter 2 Literature Review 13
2.1 Driving risk classification model 13
2.1.1 Pay-As-You-Drive model 13
2.1.2 Behavior-centric model 16
2.1.3 Driving Pattern model 19
2.2 Driving maneuver 21
2.2.1 Driving behavior 21
2.2.2 Near crash event 23
2.3 Summary 25
Chapter 3 Methodology 27
3.1 Layer 1 Raw vehicular dynamic records 28
3.2 Layer 2 Vehicular dynamic records 28
3.3 Layer 3 Driving behaviors & Near crashes 28
3.3.1 Driving behaviors & Near crashes 28
3.3.2 Driving score & Risk level 32
3.4 Layer 4 Driving pattern-N 35
3.4.1 Association Rule mining 35
3.4.2 Sequential Pattern mining 37
3.5 Layer 5 Prediction of driver’s risk level 38
3.5.1 Random forest algorithm 38
3.5.2 Driving pattern-N model 40
3.5.3 Driving pattern model 41
3.5.4 Behavior-centric model 43
3.5.5 Statistic model 43
3.5.6 Performance evaluation 44
Chapter 4 Experiments 48
4.1 Experiment 1 50
4.1.1 Experiment Design of experiment 1 50
4.1.2 Experiment results of experiment 1 50
4.1.3 Summarize of experiment 1 60
4.2 Experiment 2 61
4.2.1 Experiment Design of experiment 2 61
4.2.2 Experiment results of experiment 2 61
4.2.3 Summarize of experiment 2 74
4.3 Experiment 3 75
4.3.1 Experiment Design of experiment 3 75
4.3.2 Experiment results of experiment 3 75
4.3.3 Summarize of experiment 3 89
4.4 Discussion 90
Chapter 5 UBI platform & Premiums Calculation 93
5.1 UBI platform 93
5.2 Premiums Calculation 95
Chapter 6 Conclusions & Future works 97
6.1 Conclusions 97
6.2 Future works 98
REFERENCES 99

 
List of Tables 
Table 1 WIN-WIN-WIN situation under UBI 5
Table 2 UBI programs in Taiwan 7
Table 3 Predict variables (Paefgen et al., 2013) 13
Table 4 Model variables (Baecke and Bocca, 2017) 15
Table 5 Model performance (Baecke and Bocca, 2017) 16
Table 6 Probability of at-fault accidents on different variables. 18
Table 7 Prediction model comparison of related work 26
Table 8 Details of driving events in this research 30
Table 9 Descriptive statistic of driving events from January to March 31
Table 10 The corresponding driving behaviors between Li et al. (2017) and this work 42
Table 11 Different predicted models in this research 44
Table 12 Comparison between the 3 experiments 49
Table 13 Performance of DPN-3 model in Exp. 1 51
Table 14 Performance of DPN-6 model in Exp. 1 53
Table 15 Performance of DP-3 model in Exp. 1 54
Table 16 Performance of DP-6 model in Exp. 1 56
Table 17 Performance of BC-3 model in Exp. 1 57
Table 18 Performance of BC-6 model in Exp. 1 59
Table 19 Parameters and weighted average metrics of different models in Exp.1 60
Table 20 Performance of DPN-3 model in Exp. 2 62
Table 21 Performance of DPN-6 model in Exp. 2 64
Table 22 Performance of DP-3 model in Exp. 2 65
Table 23 Performance of DP-6 model in Exp. 2 67
Table 24 Performance of BC-3 model in Exp. 2 68
Table 25 Performance of BC-6 model in Exp. 2 70
Table 26 Performance of STA-3 model in Exp. 2 71
Table 27 Performance of STA-6 model in Exp. 2 73
Table 28 Parameters and weighted average metrics of different models in Exp.2 74
Table 29 Performance of DPN-3 model in Exp.3 76
Table 30 Performance of DPN-6 model in Exp.3 78
Table 31 Performance of DP-3 model in Exp.3 80
Table 32 Performance of DP-6 model in Exp.3 82
Table 33 Performance of BC-3 model in Exp.3 83
Table 34 Performance of BC-6 model in Exp.3 85
Table 35 Performance of STA-3 model in Exp.3 86
Table 36 Performance of STA-6 model in Exp.3 88
Table 37 Parameters and weighted average metrics of different models in Exp.3 89
Table 38 Performance of each model in the three experiments 92


List of Figures
Fig. 1 The vicious cycle in auto insurance market 2
Fig. 2 Conventional model vs. PHYD model 4
Fig. 3 Different auto insurance models (Bian et al., 2018) 7
Fig. 4 Behavior-centric model vs. Driving pattern-N model 10
Fig. 5 Research framework 12
Fig. 6 Prediction performance of different models (Paefgen et al., 2013) 14
Fig. 7 AUC of models under class skew (Paefgen et al., 2013) 14
Fig. 8 Cluster results (Guo et al., 2013) 17
Fig. 9 Transition probability of typical maneuver transition patterns in different risk group (Li et al., 2017) 20
Fig. 10 Definitions of longitudinal and lateral maneuvers on highways 21
Fig. 11 Safety domain by velocity (Eboli et al., 2016) 22
Fig. 12 Safety domain by mileage (Eboli et al., 2016) 23
Fig. 13 Data analysis pyramid 27
Fig. 14 Thresholds between each risk level of evaluation from HO-HSIN 33
Fig. 15 Thresholds between each risk level of evaluation from this work 34
Fig. 16 Process of bootstrap aggregating 39
Fig. 17 Confusion matrix 45
Fig. 18 Example of ROC curve 47
Fig. 19 The process of driving risk prediction 49
Fig. 20 ROC curve of DPN-3 model in Exp. 1 51
Fig. 21 ROC curve of DPN-6 model in Exp. 1 53
Fig. 22 ROC curve of DP-3 model in Exp. 1 55
Fig. 23 ROC curve of DP-6 model in Exp. 1 56
Fig. 24 ROC curve of BC-3 model in Exp. 1 58
Fig. 25 ROC curve of BC-6 model in Exp. 1 59
Fig. 26 Prediction performance of each model in experiment 1 60
Fig. 27 ROC curve of DPN-3 model in Exp. 2 63
Fig. 28 ROC curve of DPN-6 model in Exp. 2 64
Fig. 29 ROC curve of DP-3 model in Exp. 2 66
Fig. 30 ROC curve of DP-6 model in Exp. 2 67
Fig. 31 ROC curve of BC-3 model in Exp. 2 69
Fig. 32 ROC curve of BC-6 model in Exp. 2 70
Fig. 33 ROC curve of STA-3 model in Exp. 2 72
Fig. 34 ROC curve of STA-6 model in Exp. 2 73
Fig. 35 Prediction performance of each model in experiment 2 74
Fig. 36 ROC curve of DPN-3 model in Exp. 3 77
Fig. 37 ROC curve of DPN-6 model in Exp. 3 79
Fig. 38 ROC curve of DP-3 model in Exp. 3 80
Fig. 39 ROC curve of DP-6 model in Exp. 3 82
Fig. 40 ROC curve of BC-3 model in Exp. 3 84
Fig. 41 ROC curve of BC-6 model in Exp. 3 85
Fig. 42 ROC curve of STA-3 model in Exp. 3 87
Fig. 43 ROC curve of STA-6 model in Exp. 3 88
Fig. 44 Prediction performance of each model in experiment 3 89
Fig. 45 UBI platform proposed in this research 93
                                    

Agrawal, R., & Srikant, R. (1994, September). Fast algorithms for mining association rules. In Proc. 20th int. conf. very large data bases, VLDB (Vol. 1215, pp. 487-499).
Agrawal, R., & Srikant, R. (1995, March). Mining sequential patterns. In icde (Vol. 95, pp. 3-14).
Baecke, P., & Bocca, L. (2017). The value of vehicle telematics data in insurance risk selection processes. Decision Support Systems, 98, 69-79.
Bian, Y., Yang, C., Zhao, J. L., & Liang, L. (2018). Good drivers pay less: A study of usage-based vehicle insurance models. Transportation research part A: policy and practice, 107, 20-34.
Cheng, B., Lin, Q., Song, T., Cui, Y., Wang, L., & Kuzumaki, S. (2011). Analysis of driver brake operation in near-crash situation using naturalistic driving data. International Journal of Automotive Engineering, 2(4), 87-94.
Eboli, L., Mazzulla, G., & Pungillo, G. (2017). How to define the accident risk level of car drivers by combining objective and subjective measures of driving style. Transportation research part F: traffic psychology and behaviour, 49, 29-38.
Eboli, L., Mazzulla, G., & Pungillo, G. (2016). Combining speed and acceleration to define car users’ safe or unsafe driving behaviour. Transportation research part C: emerging technologies, 68, 113-125.
Fernández-Delgado, M., Cernadas, E., Barro, S., & Amorim, D. (2014). Do we need hundreds of classifiers to solve real world classification problems? The Journal of Machine Learning Research, 15(1), 3133-3181.
Guo, F., & Fang, Y. (2013). Individual driver risk assessment using naturalistic driving data. Accident Analysis & Prevention, 61, 3-9.
Husnjak, S., Peraković, D., Forenbacher, I., & Mumdziev, M. (2015). Telematics system in usage-based motor insurance. Procedia Engineering, 100, 816-825.
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.
Li, H., Wang, Y., Zhang, D., Zhang, M., & Chang, E. Y. (2008, October). Pfp: parallel fp-growth for query recommendation. In Proceedings of the 2008 ACM conference on Recommender systems (pp. 107-114). ACM.
Ma, Y. L., Zhu, X., Hu, X., & Chiu, Y. C. (2018). The use of context-sensitive insurance telematics data in auto insurance rate making. Transportation Research Part A: Policy and Practice, 113, 243-258.
Neale, V. L., Dingus, T. A., Klauer, S. G., Sudweeks, J., & Goodman, M. (2005). An overview of the 100-car naturalistic study and findings. National Highway Traffic Safety Administration, Paper, 5, 0400.
Pei, J., Han, J., Mortazavi-Asl, B., Pinto, H., Chen, Q., Dayal, U., & Hsu, M. C. (2001). Prefixspan: Mining sequential patterns efficiently by prefix-projected pattern growth. In Proceedings 17th international conference on data engineering (pp. 215-224). IEEE.
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.
Paefgen, J., Staake, T., & Thiesse, F. (2013). Evaluation and aggregation of pay-as-you-drive insurance rate factors: A classification analysis approach. Decision Support Systems, 56, 192-201.
Srikant, R., & Agrawal, R. (1996, March). Mining sequential patterns: Generalizations and performance improvements. In International Conference on Extending Database Technology(pp. 1-17). Springer, Berlin, Heidelberg.
Sarang Narkhede, 2018, Understanding AUC - ROC Curve, website: https://towardsdatascience.com/understanding-auc-roc-curve-68b2303cc9c5 .
Tselentis, D. I., Yannis, G., & Vlahogianni, E. I. (2017). Innovative motor insurance schemes: A review of current practices and emerging challenges. Accident Analysis & Prevention, 98, 139-148.
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.

校內：2022-09-01公開
校外：2022-09-01公開

簡易檢索 / 詳目顯示

相關論文