Article Preview
TopIntroduction
Zadeh's proposal of modeling the mechanism of human thinking with linguistic values rather than ordinary (crisp) numbers led to the introduction of fuzziness into statistical and dynamical modelling and to the development of a new class of systems called fuzzy models. Fuzzy models are capable of incorporating linguistic information naturally and conveniently. The ability to deal simultaneously both with linguistic information and numerical information in a systematic and efficient manner is one of the most important advantages of fuzzy models (Zadeh, 1965; Yen & Langari, 1999).
In fuzzy modeling, one of the most important problems is the identification of a predictive model from a set of numerical data. It is the task of identifying the parameters of a fuzzy inference system so that a desired behaviour is attained (Yager & Filev, 1994). The task of fuzzy model identification is basically based on proper generation of their structure which includes membership functions, input and output parameters and rule-base (Angelov & Buswell, 2002). This task can be performed by following the steps given below (Kumar, Bhalla & Singh, 2009):
Step 1: Initialization of the rule-base structure (antecedent part of the rules).
Step 2: Estimation of the parameters of the consequent part.
Step 3: Prediction of the output of fuzzy model through standard data sets.
Step 4: Reading of the next data sample at the next time step.
Step 5: Recursive calculation of the potential of each new data sample to influence the structure of the rule-base.
Step 6: Recursive up-date of the potentials of old centres taking into account the influence of the new data sample.
Step 7: The new data sample competes with the existing rules’ centres. Decision to modify or update the rule-base structure is taken.
The problem of fuzzy model identification can be formulated as a search and optimisation problem and it becomes very difficult to realise when the available knowledge is incomplete, and the search space is very large. The survey related to this field reveals that many hard computing as well as soft computing techniques are available in the literature to solve such problems. But the best suited way to tackle this problem is proved to be the use of soft computing approaches like Genetic algorithms, Neural networks and other Nature inspired approaches. Unlike hard computing techniques, soft computing techniques do not rely on preciseness and accuracy. They provide good enough solutions with high probability and low cost.