Big Data and Predictive Analysis in Auto Insurance

Postgraduate Thesis uoadl:1447081 707 Read counter

Unit:
Κατεύθυνση Ψηφιακά Μέσα Επικοινωνίας και Περιβάλλοντα Αλληλεπίδρασης
Library of the Faculties of Political Science and Public Administration, Communication and Mass Media Studies, Turkish and Modern Asian Studies, Sociology
Deposit date:
2015
Year:
2015
Author:
Evangelos Giannoudakos
Supervisors info:
Κωνσταντίνος Μουρλάς, Αναπληρωτής καθηγητής, Τμήμα ΕΜΜΕ, ΕΚΠΑ
Original Title:
Αξιοποίηση Δεδομένων Μεγάλου Όγκου και Προβλεπτική Ανάλυση στην Ασφάλιση Αυτοκινήτου
Languages:
Greek
Translated title:
Big Data and Predictive Analysis in Auto Insurance
Summary:
One of the purposes of this diploma thesis is to underline the significance of the
investment in Big Data and Predictive Analytics Technologies, because they can give
businesses a valuable window into valuable streams of information, such as customer
purchasing habits. As the presence of the Internet of Things (IoT) — such as
connected devices, sensors and smart machines — grows, the ability of things to
generate new types of real-time information and to actively participate in an
industry’s value stream will also grow. A company’s biggest database isn’t the
transaction, CRM, ERP or other internal database. Rather it’s the Web itself and the
world of exogenous data, now available from syndicated and open data sources.
The case study in this paper is the insurance sector and particularly, the auto
insurance sector. Since the Insurance sector relies on data, the specific case study was
not selected randomly. Especially nowadays, that the auto insurance industry is facing
an unprecedented number of challenges brought on by heavy price competition and
increased claims costs, the need for better underwriting and risk management is
greater than ever.
The main purpose of this diploma thesis is to analyze all the potentials that Big
Data can generate for the auto insurance sector and the underwriting process, by
creating new risk categories and using more parameters in calculating the risk
involved in each insurance application, in order to suggest fair and personalized
premiums. Predictive analytics find application in multiple areas such as policy risk
scoring, claim fraud detection, referral scoring and score based customer
segmentation for marketing and service purposes.
In the final section of this thesis, we analyze the design and implementation of a bigdata analytics product, named «Delta Mining», with a focus on the auto insurancesector. This product will support features, such as data cleansing, data modelling,
vectorization and predictive elements dedicated to the insurance risk management and
marketing. Also, the proposed platform aims to simplify the integration and
deployment of «live» big data analytics into business processes of insurance
companies. The proposed platform will provide a complete collaborative predictive
analytics solution, which will enable the existing company's IT personnel to build
predictive models, to score those models with huge data sets and help existing
company's IT personnel with low expertise in data analytics, to understand
fundamental concepts, such as object vectorization, and thus effectively perform tasks
such as the preparation of data, apply different machine learning algorithms for
predictive analysis and train classification models.
The goal is to simplify the application of complex machine learning, pattern
recognition and data mining techniques in risk management, implementing a platform
that helps insurance organizations to estimate risk, thus having competitive
advantages in terms of cost, customer relationships (customer satisfaction), and
market leadership.
Main subject category:
Social, Political and Economic sciences
Keywords:
big data, predictive analytics, car insurance, telematics, business intelligence
Index:
No
Number of index pages:
0
Contains images:
Yes
Number of references:
168
Number of pages:
197
Notes:
Τοπική Ψηφιακή Βιβλιοθήκη
File:
File access is restricted.

ΓΙΑΝΝΟΥΔΑΚΟΣ ΕΥΑΓΓΕΛΟΣ .pdf
14 MB
File access is restricted.