Author : Banghee So
Publisher :
ISBN 13 :
Total Pages : 0 pages
Book Rating : 4.:/5 (133 download)
Book Synopsis Actuarial Models for Understanding Driver Behavior with Telematics Data by : Banghee So
Download or read book Actuarial Models for Understanding Driver Behavior with Telematics Data written by Banghee So and published by . This book was released on 2021 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Powered with telematics technology, insurers can now capture a wide range of data to better decode driver's behavior, such as distance traveled and how drivers brake, accelerate or make turns. Such additional information helps insurers improve risk assessments for usage-based insurance (UBI), an increasingly popular industry innovation. In this thesis, we first explore how to integrate telematics information to improve understanding of driver heterogeneity, as well as to better predict accident counts. For motor insurance during a policy year, we typically observe a large proportion of drivers with zero accidents, a less proportion with exactly one accident, and far fewer with two or more accidents. We introduce the use of a cost-sensitive multi-class adaptive boosting algorithm, which we call SAMME.C2, to handle such imbalances in a classification model. Using the SAMME.C2 algorithm, we find improved assessment of driving behavior with telematics relative to traditional risk variables. We next demonstrate the theoretical justification of the SAMME.C2 algorithm in two respects: (1) it is equivalent to Forward Stagewise Additive Modeling with exponential loss, and (2) it is a Bayes classifier. When cost-sensitive learning is added, we find the superiority of SAMME.C2 in controlling for issues related to class imbalances, especially when compared to just the SAMME algorithm. We performed numerical experiments to better understand the distinguishing characteristics of the algorithm. Finally, this thesis describes the techniques employed in the production of a synthetic dataset of driver telematics that is emulated from a real insurance dataset. The method uses a three-stage process that involves deploying machine learning algorithms. It is aimed to produce a resource that can be used to advance models to assess risks for usage-based insurance. It is the hope of this work to provide and encourage the research community to explore innovative methods relevant to such data. The synthetic dataset produced includes 100,000 observations about driver's claims experience (both claim counts and amounts were generated) together with associated classical risk variables and telematics-related variables. We further show, using visualization, model fitting, and data summarization, how remarkable the similarities are between the synthetic and the real datasets.