Life Insurance Risk Assessment

This application illustrates the use of Togai's upcoming FuzzyCLIPS product to 
build a knowledge based decision support system possessing fuzzy components to 
improve knowledge base system performance.  The results employing FuzzyCLIPS 
are compared with the results obtained from the solution of the problem using 
traditional numerical equations.  The design of the fuzzy solution consists of 
a CLIPS rule-based system for some factors combined with fuzzy logic rules for 
others.  

In addition, the TILShell is a quick and effective way of designing a rulebase 
with a graphical interface.  TILShell is  an excellent tool for verification of 
expert system behavior and fuzzy knowledge based systems with the simulation 
capabilities available.  TILShell is also useful for tuning membership 
functions.  (See TILShell 3.0 article on page 1 for a description of new 
features.) 

Problem Statement

An insurance company needs to assess the degree of health risk associated  with 
each client based on physical characteristics, such as height, weight, and age, 
and habits related to exercising, smoking, drinking, and eating.  The output 
risk value serves as the basis for insurance premiums billed  to clients.
Those premiums have a base rate (perfect health, good habits, 35 years old) and 
an increment to adjust the premium based on the risk.  A system that produces a 
risk value between a determined range suffices to set a net rate. The equation 
is

    Cost to Insure Client = Base Rate + ((Risk /Base Risk)-1)*Increment

Implementation Strategy

The input variables of the system are of two different types: base and
incremental.  The base type of input variables  are Age, Weight, and Height.
A derived internal variable is the body mass index (BMI) that estimates fitness 
or body fat content.  Incremental input variables deal with particular habits 
and characteristics of prospective clients: exercising, dairy products intake, 
red meat intake, vegetable intake,  fat/sweet intake, smoking, and drinking 
alcoholic beverages.  The outputs of the system are the degree of risk and the 
amount of the annual premium.  The system also produces a truth value
associated with each output fuzzy set, i.e., the degree to which each fuzzy set 
defining risk contributes to the output value of risk.

The relation between decision factors and the rate change need be neither 
incremental nor linear, i.e., separate consideration of the decision factors
may not determine a change in rate that can be simply summed to determine the 
net rate.  Complex non-linearity and interdependence of the factors mean that 
computer-based decision aids are useful to a human agent and that sharp
decision boundaries such as those produced by a normal rule-based system are 
sensitive to small uncertainties in the input data.  Fuzzy logic provides a 
basis for accommodating such uncertainty with finesse.

For this particular example, five different sets of fuzzy rules are defined. 
Figure 3 shows the structures.  The first rulebase relates a risk_1 to age and 
BMI.  The second rulebase relates a risk_2 to smoking and drinking habits. The 
third rulebase relates a risk_3 to the amount of exercise and intake of 
vegetables.  The last rulebase relates a risk_4 to intake of dairy products,
red meat, and fat and sweet products.  A fifth rulebase relates risks 1-4 to
the overall risk to complete the risk assessment.  The importance of breaking 
down the problem into smaller related groups is the fact that the number of 
rules needed to control the system decreases dramatically. In our example, the 
number went from 8748 possible rules  to a maximum of 313 rules (only 89 are 
used in this implementation) by this problem decomposition.

User Interface

The user interface allows the user to choose values for frequency and intensity 
of incremental factors against  an arbitrary scale, introducing the potential
to fuzzify the input to conduct reasoning with correlation and interpolation 
between benchmarks or way points.  The system can also be implemented to work
in batch mode to read large amounts of data from a file.

Results

Experiments were implemented to test the system.  The results of the fuzzy
logic system were compared with the results of an expert system using 
traditional methods.  A sample of the results are shown in Figure 4 to 
illustrate that risk evaluation using fuzzy logic avoids artificial jumps at 
discrete boundaries.  The parameters used were constant weight/height 
relationship, constant eating and exercise habits, and variable age.

In summary, the fuzzy system provides advantages for insurance companies and 
insurance clients: behavior of the system can be controlled, predictability of 
risk measures and premium amounts, better financial forecasting, fair 
distribution of risk/premiums among population, and predictability of future 
premiums among others.

As has been demonstrated with the life insurance risk assessment system, 
FuzzyCLIPS  combined with the TILShell can be used to implement traditional 
business applications effectively.  These tools combine fuzzy processing with 
rule-based processing and provides improved decision aid for evaluating risk
for life insurance.   FuzzyCLIPS will be available from Togai InfraLogic this 
summer.  TILShell is available immediately.  Please call for more information
or a free demonstration disk.
