Telematics and Contextual Data Analysis and Driving Risk Prediction
| Main Author: | |
|---|---|
| Language: | English |
| Published: |
The Ohio State University / OhioLINK
2020
|
| Subjects: | |
| Online Access: | http://rave.ohiolink.edu/etdc/view?acc_num=osu1577032799045121 |
| id |
ndltd-OhioLink-oai-etd.ohiolink.edu-osu1577032799045121 |
|---|---|
| record_format |
oai_dc |
| collection |
NDLTD |
| language |
English |
| sources |
NDLTD |
| topic |
Computer Science Computer Engineering Telematics Data Analysis Driving Behavior Driving Risk Usage Based Insurance Urban Planning Traffic Management Accident Prediction Traffic Prediction |
| spellingShingle |
Computer Science Computer Engineering Telematics Data Analysis Driving Behavior Driving Risk Usage Based Insurance Urban Planning Traffic Management Accident Prediction Traffic Prediction MoosaviNejadDaryakenari, SeyedSobhan Telematics and Contextual Data Analysis and Driving Risk Prediction |
| author |
MoosaviNejadDaryakenari, SeyedSobhan |
| author_facet |
MoosaviNejadDaryakenari, SeyedSobhan |
| author_sort |
MoosaviNejadDaryakenari, SeyedSobhan |
| title |
Telematics and Contextual Data Analysis and Driving Risk Prediction |
| title_short |
Telematics and Contextual Data Analysis and Driving Risk Prediction |
| title_full |
Telematics and Contextual Data Analysis and Driving Risk Prediction |
| title_fullStr |
Telematics and Contextual Data Analysis and Driving Risk Prediction |
| title_full_unstemmed |
Telematics and Contextual Data Analysis and Driving Risk Prediction |
| title_sort |
telematics and contextual data analysis and driving risk prediction |
| publisher |
The Ohio State University / OhioLINK |
| publishDate |
2020 |
| url |
http://rave.ohiolink.edu/etdc/view?acc_num=osu1577032799045121 |
| work_keys_str_mv |
AT moosavinejaddaryakenariseyedsobhan telematicsandcontextualdataanalysisanddrivingriskprediction |
| _version_ |
1719456731790049280 |
| spelling |
ndltd-OhioLink-oai-etd.ohiolink.edu-osu15770327990451212021-08-03T07:13:36Z Telematics and Contextual Data Analysis and Driving Risk Prediction MoosaviNejadDaryakenari, SeyedSobhan Computer Science Computer Engineering Telematics Data Analysis Driving Behavior Driving Risk Usage Based Insurance Urban Planning Traffic Management Accident Prediction Traffic Prediction Analysis of telematics data collected from drivers in real-time, along with contextual data (such as traffic and weather, data), provides valuable insights regarding an individual's driving behavior, common driving habits, and characteristics of a road network. The primary focus in this dissertation is on predicting the risk in driving, with the risk being a combination of the risk of the driver, the risk in a route, and the risk caused by driving conditions. We propose several data-analytic techniques for driving risk prediction and related caused and apply them to different sources of telematics and contextual data to extract useful insights. We seek to answer the following research question: How can telematics data and their context be modeled to make a fair and sound prediction about driving risk? In answering this question, we take two concerns into account: 1) scoring of driving risk is not a context-agnostic process, and 2) driving risk is not independent of the personality of drivers and their driving skills. We propose a solution that consists of three parts: a) characterizing driving context, b) characterizing driving style, and c) context-aware driving risk prediction. The first two parts derive useful insights that we leverage to design the third part. Characterizing driving context is about exploring properties of different contexts. We propose two solutions for this task. The first solution, which we term segmentation and causality analysis, derives the characteristics of contexts from the aggregate behavior of drivers. We start by segmenting trajectories to identify meaningful driving patterns (e.g., a hard brake). Then we analyze each pattern with respect to contextual data to identify cause-and-effect patterns (e.g., traffic signal --> hard-braking event). The second solution is a geo-spatiotemporal pattern discovery framework based on contextual data. We propose a new framework that explores two types of patterns, propagation and influential. Propagation patterns show common cascading patterns of entities (e.g., rain --> accident --> congestion). Influential patterns show the impact of long-term entities on their spatial neighborhood (e.g., major construction --> more congestion events). Characterizing driving style is about illustrating drivers' personalities and skills, by capturing variations in driving behavior that discriminate different drivers from each other. We propose a deep-neural-neural-network model to derive useful driving style information from telematics data. Context-aware driving risk prediction can be achieved by macro- and micro-level solutions. Macro-level driving risk prediction is about predicting the possibility of traffic accidents inside a region during a time interval. To accomplish this, we propose a novel deep-neural-network-based solution that uses contextual data (such as traffic, weather, characteristics of road-network, and temporal data). Micro-level driving risk prediction is about utilizing telematics and contextual data to perform individual-level risk prediction. We propose a new framework that relies on telematics and contextual and it includes: 1) contextualizing telematics data to better represent driving behavior in different contexts; 2) building a risk cohort classifier using contextualized telematics data and weak risk labels; and 3) constructing a driving risk prediction process to be employed for real-world purposes. 2020-09-25 English text The Ohio State University / OhioLINK http://rave.ohiolink.edu/etdc/view?acc_num=osu1577032799045121 http://rave.ohiolink.edu/etdc/view?acc_num=osu1577032799045121 unrestricted This thesis or dissertation is protected by copyright: some rights reserved. It is licensed for use under a Creative Commons license. Specific terms and permissions are available from this document's record in the OhioLINK ETD Center. |