Detecting Effectiveness of Outliers and Noisy Data on Fuzzy System Using FCM

Fuzzy systems which are an artificial intelligent technique are applicable for controlling and decision support systems. Fuzzy systems are created using membership functions (MFs) which modeled based on dataset. Therefore, there is relation between uncertainty of input data and fuzziness expressed by MFs. Outliers and noisy data are kinds of uncertainty which affect on membership function. Thus, MFs will not be a robustness model to make an accurate decision for controlling and decision making. However, the isolation of outliers is important both for improving the quality of original data and for reducing the impact of outlying value in processes of knowledge discovery and MFs. In this paper, the statistic equation is applied to detect and remove noisy data, and for detecting outliers; clustering techniques are utilized to group the outlying value based on their similarities. From the huge number of clustering methods Fuzzy C-Mean clustering is a strong tool for the identification of changing class structures and it is flexible, moveable and creatable for uncertain data (i.e. Outlier) in order to improve the accuracy in fuzzy systems. The outlier detection procedure is demonstrated on weather dataset