Data Mining in Insurance Industry |
 |
INSURANCE AND KNOWLEDGE DISCOVERY
Companies in the Insurance industry collected enormous amounts of data
about their clients. This is invaluable information about customers behaviour,
activities, and preferences.
To extract information from the whole amount of raw data Insurance firms
lost time and efforts, due to their protective regulations.
The data extracting process resulted in developing new products and services
to meet customers needs. It unearthed information on customers, markets, and
competitors. But some essential dependencies and patterns just cannot be
discovered by a human in terabytes of data, while increasing the competition
made Insurance companies to become more effective and customer-centric.
That is when Insurance industry needed Data Mining Software as a new
tool for
knowledge discovery.
EXAMPLE OF DATA MINING IN INSURANCE INDUSTRY
- The first step for an analyst is to select databases (data warehouses)
that can be used for knowledge discovery.
- After selecting an appropriate database for data mining it is necessary to
set up the data mining software. The most suitable data mining method
for fraud detection in insurance is Classification. Data mining software
identifies groups of customers with distinct behaviour patterns.
- The next step is “training” the data mining software. During this
step the program will find decision rules, patterns, behaviours, trends
and deviations from the norm.
After the knowledge discovery you will obtain decision rules or decision
trees - a
prediction model, containing structured features of insurance frauds or loyal
clients. The decision rules might be used for predicting which
customers or potential customers might commit fraud, and for further
knowledge discovery.
EXAMPLE OF DATA MINING IN INSURANCE. STEP-BY-STEP GUIDE
Creating Decision Rules:
- Select clients database
- Select a table
- Make sure the table contains records of fraudulent clients
- Make sure the table contains a field that marks fraudulent clients
- Open the database and the table with the help of
Estard Data Miner wizard
- Start the query wizard
- Select the field that marks fraudulent clients as the Class examined field
- View classes that where determined by Estard Data Miner
- Select Rules
- Set Rules creating options. For example you can set the minimum number of
records that should be described by the Rule. This means that if you select
ten, Rules that describe only nine records, or less, will not be displayed
by the program.
- Start the query by pressing the Finish button.
- The created Rules will be automatically displayed on the corresponding
program page.
Now the received Rules can be used to determine whether the client is
likely to be fraudulent.
Another way of using the Decision Rules is to determine possible fraudulent clients
in existing databases.
This can be performed on the
Analysis page.
To select fraudulent clients from database follow these steps:
- Select a clients database to be analysed
- Select a table to be analysed
- Open the database and the table with the help of
Estard Data Miner wizard
on the
Analysed Data Set page
-
Select the clients class or Decision Rules that will be used for
searching in the database and press the "Process Query" button in the same
window (you should previously create Decision Rules)
- View the results of searching on the
Analysed Data Set page
HOW DATA MINING WILL RESULT?
For example, you can use data mining software to predict which customers will buy new policies.
You also can :
-
Spot and understand new business trends in claims
- Predict and detect fraudulent or risky behaviour: estimate high risk customers.
-
Detect customers, belonging to frequent buying patterns.
|
|
|
24.07.08
