Abstract
Content
- Introduction
- 1. Theme urgency
- 2. The purpose and tasks of the Master's work
- 3. Selection methods for solving problems
- Conclusion
- References
Introduction
Modern insurance companies provide a wide range of vehicle insurance, and customers are not always easy to sort out all the existing proposals and to select the most appropriate type of insurance to them.
This paper is a description of the development of the prototype expert system that aims to help in determining the best insurance option. This system is intended to replace an expert consultant in the chosen subject area.
1. Theme urgency
With the increasing technical capabilities of modern computer systems, more and more possibilities for data processing. Technological progress has not bypassed the insurance company. Many foreign companies use a lot of complex scoring systems. The urgency is due to the increased demand Ukrainian insurance companies in effectively underwriting in connection with the development of the insurance market and virtually complete absence of embedded scoring systems.
2. The purpose and tasks of the Master's work
Objective: Development of a computerized scoring system to support decision-making car insurance.
Problems:
- Investigation of underwriting
- Study of factors influencing the choice of the insurance portfolio
- Finding the optimal insurance portfolio
- Investigation of methods for predicting insurance claims
- Selection of the optimal method for creating ES
3. Selection methods for solving problems
Since the main objective of the development of the DSS is to select the best insurance package, I believe neylorovskie diagnosing logical to use the system, because thanks to this structure, we can be certain probabilities to determine the suitability of the insurance package for each car owner.
We use the method of Naylor to our area and form a knowledge base for decision support system model car insurance.
Hypotheses Hi:
The insurance package 1, 0.1, 5, (1, 0, 0.99), (2, 0.7, 0.05), (4, 0.2, 0.5), (5, 0, 0.99 ), (6, 1, 0.01);
The insurance package 2; 0,05; 2; (2; 1; 0,01); (6; 0,9; 0,02);
The insurance package 3; 0,01; 3; (3; 0,9; 0,1); (4; 0,25; 0,5); (6; 0,9; 0,02);
The insurance package 4; 0,01; 2; (4; 0,01; 0,5); (6; 0,9; 0,02).
We give a transcript of the above records. This probability that the owner of auto insurance will suit package 1 is equal to 0.1. This knowledge base with two related insurance package two (2) criteria.
The first criterion - the highest sum insured. High probability of the insured amount is set equal to 1. The probability that a high amount of this insurance package does not fit is 0.01.
The second criterion - the brand model. Probability expensive brand when choosing a car insurance package 1 is equal to 0.9. The likelihood that the expensive brand cars this package does not fit is 0.02.
Evidence of Ei:
- The number of insurance claims, many insurance claims?
- The insurance amount; exceeds the sum insured?
- Car - the frequency of hijackings, theft rate?
- Year of manufacture cars, new car?
- Country of origin car; Country of cars?
- Car - the frequency of accidents, road car?
Table 1 – Prices evidence
According to the calculated initial rates of certificates C1 (Ei) is always the first to ask a question related to the certificate of E6 (it has the highest price of 2.2381)
Expensive car?
When you answer YES member (R = 5), we can count the array P (Hi) and calculate the new prices evidence C2 (Ei). With the new price of certificates will be asked the following question related to the testimony of E2 (it has the highest price of 1.2135).
The message is that the car is expensive, resulting in a substantial increase in the price and other evidence that looks very natural.
Conclusion
The methods of creating DSS car insurance. Chose an optimal method. On the basis of selection gave an example of finding a solution. In the future we plan to study to determine the real "weight" of all the factors, and choose the programming environment of the DSS.
Important
In writing this abstract master's work is not yet complete. Final completion - December 2013. The full text of the work and materials on the topic can be obtained from the author or his head after that date.