Christ University Institutional Repository

SVM Ensemble for Insurance Data Analysis

T. MATHEW, SIJI (2012) SVM Ensemble for Insurance Data Analysis. UNSPECIFIED thesis, UNSPECIFIED.

[img]PDF
Restricted to Registered users only

1900Kb

Abstract

Data mining is the process of analysing data from different perspectives and summarizing it into useful information. Companies with a strong consumer focus use data mining. The information getting from datamining is useful to increase revenue and reduce overall costs of the company. It is applied in retail field, financial sector, communication media, and in marketing organizations. Datamining facilitate these companies to determine relationships among company internal factors such as price, product positioning, or staff skills, and external factors such as competition in products, economic indicators, and customer demographics. Ensemble learning is a machine-learning paradigm where multiple models or learners are trained to solve the problem. This research explores the usage of SVM ensemble for Insurance Data Analysis. The number of Insurance firms is increasing day by day. The main objective of this research is to find out the best policy from a given list of Insurance policies. In this research a detailed study of SVM ensemble is done. An insurance dataset obtained from UCI knowledge discovery in Databases Archive is taken in the research analysis. From the dataset five different Non-life insurance policies were selected and used in this research work. The categories of policies include Fire policy, Home policy, Car policy, Kissan policy and Boat policy. AdaBoost, multiclassifier SVM ensemble was created and tested with the insurance dataset. SVM ensemble produces better accuracy than other ensembles. The knowledge flow of SVM ensemble is loaded in Weka. From each category, the policy that gives a highest accuracy value for SVM ensemble is considered to be the best policy. A graphical user interface is also developed using .NET framework, to view the policy output. This system helps the user to find out a best policy from the analysed data. KEYWORDS: SVM ensemble, Insurance policy, Accuracy, ROC, Support Vector Machine.

Item Type:Thesis (UNSPECIFIED)
Subjects:Thesis > MPhil > Computer Science
Divisions:M Phil > Computer Science
ID Code:4983
Deposited By:Knowledge Center Christ University
Deposited On:21 Oct 2013 12:46
Last Modified:31 Oct 2013 15:08

Repository Staff Only: item control page