Fuzzy clusters represent the data's degree of pertinence in function of its distance from the data groups that have some similarity. Fuzzy clusters are widely used to infer fuzzy rules from the data. The evolving algorithms has gained notoriety due to its learning capacity and performance in on-line applications. This paper intends to do a further analysis on the performance of evolving algorithms in comparison to other classical approaches.
Fuzzy clusters represent the data's degree of pertinence in function of its distance from the data groups that have some similarity. Fuzzy clusters are widely used to infer fuzzy rules from the data. The evolving algorithms has gained notoriety due to its learning capacity and performance in on-line applications. This paper intends to do a further analysis on the performance of evolving algorithms in comparison to other classical approaches.
Fuzzy clusters represent the data's degree of pertinence in function of its distance from the data groups that have some similarity. Fuzzy clusters are widely used to infer fuzzy rules from the data. The evolving algorithms has gained notoriety due to its learning capacity and performance in on-line applications. This paper intends to do a further analysis on the performance of evolving algorithms in comparison to other classical approaches.
distance from the data groups that have some similarity. This structures are widely used to infer fuzzy rules from the data, which is necessary when the system's complexity exceeds the capability of the human expert to infer a sufficient number of rules that describes the cluster's influence in the output of the system. From this premise, in the context of training a clustering algorithm, each sample's pertinence to a cluster, and the associated output are a pair of characteristic, so the functions that can be estimated from these pairs are rules that describes the influence of that particular cluster on the system's output. Among the algorithms used to obtain fuzzy cluster, the evolving algorithms has gained notoriety due to its learning capacity and performance in on-line applications. This paper intends to do a further analysis on the performance of evolving algorithms in comparison to other classical approaches using modern metrics to validate clusters, other than those that were applied to validate the results at the time they were published. These modern metrics have shown better precision in estimate how well the clustering algorithm can make consistent clusters, without being affected by variations solely due to change in the parameters, as occurs with indices like partition coefficient, classification entropy and proportion exponent.