Presenting a model to identify the influencing factors and their impact factor in the profit and loss of the third party car insurance of insurance companies by means of data mining methods, a case study of Iran Insurance Company.

Master's Thesis

in the field of Computer - Software Engineering

Abstract

Reviewing car insurance information has shown factors such as the type of car used, having a driver's license, the type of license and whether or not it matches the vehicle, the amount of the insurance premium, the amount of insurance policy obligations, the quality of the car manufacturers, the age of the driver, the literacy of the driver, the mismatch of the insurance premium with the insurance case, the delay in renewing the insurance policy, the profit and loss of insurance companies. have had an impact.

The aim of this thesis is to know the influencing factors and their coefficient of influence on the profit and loss of third party car insurance of insurance companies using data mining method and then choosing the algorithm that has the best prediction accuracy to detect these factors. Acceptable clusters will be able to provide a model to identify the influencing factors and determine the extent of their effect on the profit and loss of the third party car insurance policy.

Keywords: data mining of the third party car insurance profit and loss

Chapter 1

Introduction

Commercial and commercial companies must always focus on making profits and reducing losses in order to survive and maintain the market. and their losses are emphasized, therefore customer attraction methods as well as techniques to prevent or reduce losses are at the top of the agenda of these companies.

Among the companies that are exposed to reduced profits or increased losses due to various reasons are insurance companies. Factors such as marketing, customer loyalty, insurance premium rates, advertising, fraud, can attract or repel customers, which has a direct and indirect effect on profit and loss.

Payment of damages, as an obligation of insurance companies, leads to a decrease in profit and in some cases causes losses to an insurance company. Damage can occur for various reasons, or another action can be presented as damage, which is not the case [Derrig et. al 2006].

Factors such as driving culture, having a driver's license, the type of license and its compatibility or non-compatibility with the vehicle, intercity roads and inner-city streets that challenge municipalities and road departments, fraud, weather conditions, car quality of car manufacturers, driver's age, driver's literacy, mismatch of insurance premium with the insurance item [Wilson 2003], holidays, trips and many other things can cause damage and ultimately increase the loss of a company. get insurance.

Industrial insurance is useful, necessary and effective in economic development. Due to "increasing security in various areas of life and economic activities", "increasing investment and employment and economic growth" and "promoting economic justice and reducing poverty caused by risks", this industry has an important place in the progress and excellence of a country.

Despite the important role of insurance in creating a foundation and providing favorable economic conditions, the current state of this industry in the national economy is far from its ideal state. The general lack of familiarity and low demand for insurance products, low technical knowledge in the field of insurance services, mismatch of risk with insurance premiums, significant difference in the comparison of the risk detection criteria of third-party insurance with the equivalent type of insurance in developed countries, the existence of inadequacies in the management of insurance supply units are among the reasons for the lack of proper development of this industry in the country. Since mankind has achieved many salvations and successes throughout history with the help of science and experience, a more scientific look at the problems of this industry and finding a solution in the context of science can be the solution. Today, by data mining methods, the relationship between various effective or ineffective factors in a subject is determined, and considering that data mining is a useful tool in extracting knowledge from mass data that shows the hidden connections between them, commercial companies are turning to these techniques.

Data mining is not limited to the use of technologies and will use whatever is useful for it. However, statistics and computer are the most widely used sciences and technologies used in data mining. 1-1 Definition of data mining Data mining is the process of discovering unknown and useful rules and knowledge from a large amount of data and databases [Liu et. al 2012].

Performing data mining, like any other operation, has its own steps which are as follows:

1-Separation of useful data from extraneous data

2-Integration of different data under a single format

3-Selecting the necessary data from among other data

4- Transferring data to the data mining environment to discover rules

5-Creating related models and patterns by Data mining methods

6-Evaluating the model and patterns created to determine their usefulness

7-Dissemination of extracted knowledge to end users

1-2         Definition of insurance

Insurance: insurance is a contract by which one party undertakes to pay money or funds from the other party in case of an accident or damage caused to him. compensate or pay a certain amount. The obligor is the insurer, the party to the obligation, the insurer, the amount paid by the insurer to the insurer, the insurance premium and what is insured is the subject of insurance [Article 1 of the Insurance Law approved on 7/2/1316]. Insurance companies and check their effectiveness. Algorithms used in this research included categories, clusters, decision trees and association rules.

1-4 steps of research

In this thesis, using data mining methods, a part of the insurance company's issue and loss data for one year is modeled and a pattern is made from them. In fact, in this way, the algorithm is taught that the connections between the data lead to what results. Then a part of the data that was not used in the previous step is given to the created model and the results are evaluated by scientific criteria. In order to test the performance, other data can be given to the model and the results can be compared with the actual results.

1-5 Thesis structure

This thesis will consist of four chapters, the first chapter contains an introduction and the necessity of the research and the purpose of this research. In the second chapter, some data mining techniques and their methods are presented and the researches that have been done in this field are examined. In the third chapter, we will give a detailed description of the research done and the data mining software used in this thesis, and with the help of data mining techniques, models will be presented and the models presented in each group will be compared with each other and the best model will be selected among them. In the fourth chapter, the raised issues will be summarized and the results will be presented, and then the changes that can be made in this field in the future will be suggested. Chapter Two: Literature and Previous Researches: In this chapter, we will first have an overview of data mining methods, then we will review previous researches. 2-1: Data mining. and machine learning

Data mining is a combination of machine learning techniques, pattern recognition, statistics, database theory and summarizing and relating interesting concepts and patterns automatically from large corporate databases. The main purpose of data mining is to help the decision-making process by extracting knowledge from data [Alpaydin 2010]. The purpose of data mining is to reveal trends or patterns that have been unknown so far to make better decisions, which achieves this goal by applying statistical methods such as logistic analysis and clustering, as well as by using data analysis methods obtained from other disciplines (such as neural networks in artificial intelligence and decision trees in machine learning) [Koh & Gervis 2010. Data mining is a useful tool for exploring knowledge from large data because it predicts future trends and behaviors by observing the hidden patterns of the organization. Data mining is finding meaningful information from a large number of data by some technologies as a procedure to discover knowledge from the database, the steps of which include the following [Han and Kamber 2001].