Abstract—One of the main problems when handling the real world problems is the uncertainty degree of input data. Uncertainty factor can be a result of random variables existence, incomplete or inaccurate data, and approximations instead of measurements or incomparability of data (resulting from varying measurement or observation conditions). Interval and fuzzy numbers generally use for representation of real data. There are two main innovations in this paper: I) Classification of real data using fisher discriminator (FD), and II) Quadratic programming of FD problem with fuzzy parameters has led us to a quadratic fuzzy objective function and quadratic fuzzy constraints, that is solved for the first time in this paper. The proposed Fuzzy FD (FFD) obtain new version of classifier with two new points. I) Three region of decision are given include class 1, class 2, outlier class. II) We can classify real data with given uncertainty degree. Experimental results are performed over and Power of the FFD is seen nicely.
Index Terms—Fisher discriminator; Fuzzy data; Fuzzyquadratic programming problems; Real data classification.
Boshra Rajaei (corresponding author) is with Computer Department of Ferdowsi University of Mashhad, Iran (phone: +985115028601; email: firstname.lastname@example.org).
Hadi Sadoghi Yazdi is with Computer Department of Ferdowsi University of Mashhad, Iran (email: email@example.com, firstname.lastname@example.org).
Sohrab Effati is with Mathematics Department of Ferdowsi University of Mashhad (email: email@example.com, firstname.lastname@example.org).
Cite: B. Rajaei, H. Sadoghi Yazdi, and S. Effati, "Fisher over Fuzzy Samples," International Journal of Computer Theory and Engineering vol. 1, no. 5, pp. 632-637, 2009.