4.2. Fuzzy Logic based model
A Sugeno-type fuzzy logic model was proposed to develop using the Fuzzy Logic Toolbox of MATLAB®. For
extracting the fuzzy classification rules from the data itself clustering techniques was used. A popular
unsupervised clustering technique is the subtractive clustering technique. For this, genfis2 function in MATLAB,
which uses subtractive clustering algorithm for identifying cluster centers and radii, was chosen to build the fuzzy
inference system. According to Bataineh (2011) subtractive clustering technique outperforms other clustering
techniques like fuzzy c-means clustering as latter requires training algorithm and produces inconsistent results for
different runs. It was observed by Chopra et. al. (2004) that use of subtractive clustering reduces number of rules
than that framed in fuzzy logic system without clustering. Thus a series of experimental runs were performed to
ascertain the optimum radius of clusters and consequently the cluster radius was kept as 0.7.
Data were divided into two sets - training and testing datasets in a ratio of 85:15. Input variables used in this
model remained the same as used for the MNL model. Binary valued choice was considered as the output
variable where 0 signifies that a particular mode had not been chosen and 1 otherwise. About 95 rules were
obtained using the genfis2 function in MATLAB as shown below: