Readme_first.txt

Before running HABFUZZ, note that it has pre-defined fuzzy sets
for both the predictor and response variables.

Since version 2.8, there are two versions available:
1. The classic HABFUZZ:
Input 1: Fuzzy input with 5 membership functions (either trapezoidal or triangular)
Input 2: Fuzzy input with 5 membership functions (either trapezoidal or triangular)
Input 3: Crisp input with 8 classes
Input 4: Fuzzy input with 5 membership functions (either trapezoidal or triangular)
Output: Fuzzy output with 5 membership functions (ranging from 0 to 1)

2. The full fuzzy HABFUZZ:
Input 1: Fuzzy input with 5 membership functions (either trapezoidal or triangular)
Input 2: Fuzzy input with 5 membership functions (either trapezoidal or triangular)
Input 3: Fuzzy input with 5 membership functions (either trapezoidal or triangular)
Input 4: Fuzzy input with 5 membership functions (either trapezoidal or triangular)
Output: Fuzzy output with 5 membership functions (ranging from 0 to 1)

To run the software for your data, you may need to appropriately
modify each fuzzy set (through the STEERING files).

For example, the response variable is set to a 0-1 range and its relevant
parameters are set in the fdeclarations.f95 file as
ka=0.2, kb=0.4, kc=0.6, kd=0.8 to define five classes and
eua=0.1, eub=0.3, euc=0.5, eud=0.7, eue=0.9
for the expected utility function of the fuzzy rule-based Bayesian algorithm.

If your response variable ranges from 0 to 1000 for example,
the easiest way would be to normalize it by
dividing with the maximum value observed and run HABFUZZ or
you could change the relevant parameters, that is,
ka=200, kb=400, kc=600, kd=800
eua=100, eub=300, euc=500, eud=700, eue=900,
recompile and run.

Note that for response variables outside the 0-1 range, only the fuzzy
rule-based Bayesian algorithm is currently working.