Discussion of “Hybrid Soft Computing Approach for Mining of Complex Construction Databases” by Wen-Der Yu

Abstract

The innovative soft computing work by Yu 2007 for mining complex construction databases is appreciated, but some misuses and flaws occur in their basic conception and application of the fuzzy logic FL theory. Hybridized soft computing techniques might bring forth results that may not be in agreement with the physical and logical principles of FL Sen 2006, 2007 . To avoid such pitfalls, it is advised to concentrate on the basic FL principles prior to modeling and to make physical, plausible, and logical interpretations after the training of the model. Otherwise, complete dependence on hybrid software may lead only to erroneous scientific conclusions. The discussers would like to present the following important points, which the potential fuzzy modelers should take into account in future researches. 1. Labeling each of the fuzzy subsets with words such as low, medium, and high is a fundamental process in fuzzification Sen 2004 . However, such labels are not assigned to fuzzy subsets in the study, which makes it difficult to understand the physical basis of the model and to interpret the numerical results accordingly. 2. In FL philosophy, most left and right membership functions MFs should reach membership degrees MDs equal to 1 Sen 2004, 2007 . Unfortunately, in Fig. 1 the rightmost MF for the output variable is encircled to show the decrease toward high values. Similarly, the leftmost MFs for the input and output variable portions in rectangular blocks decrease toward low values. These locations are linguistically incorrect in terms of FL philosophy. Of course, similar illogical