A class of fuzzy clustering algorithms based on a recently introduced 'noise cluster' concepts is proposed. A 'noise prototype' is defined such that it is equi-distant to all the points in the data-set. This allows for detection of clusters amongst data with or without noise. It is shown that this concept is applicable to all the generalizations of fuzzy or hard k-means algorithms. Various applications are also considered. Application of this concept to a variety of regression problems is also considered. It is shown that the results of this approach are comparable to many robust regression techniques. The paper concludes with a summary and directions for future work.