A Proposed Approach for Solving Rough Bi-Level. Programming Problems by Genetic Algorithm

Similar documents
Rough Connected Topologized. Approximation Spaces

A Solution Methodology of Bi-Level Linear Programming Based on Genetic Algorithm

Some Inequalities Involving Fuzzy Complex Numbers

Math 1314 Lesson 24 Maxima and Minima of Functions of Several Variables

KANGAL REPORT

Concavity. Notice the location of the tangents to each type of curve.

Mobile Robot Static Path Planning Based on Genetic Simulated Annealing Algorithm

A new approach for ranking trapezoidal vague numbers by using SAW method

Larger K-maps. So far we have only discussed 2 and 3-variable K-maps. We can now create a 4-variable map in the

Piecewise polynomial interpolation

WHEN DO THREE LONGEST PATHS HAVE A COMMON VERTEX?

ROUGH MEMBERSHIP FUNCTIONS: A TOOL FOR REASONING WITH UNCERTAINTY

Intelligent Optimization Methods for High-Dimensional Data Classification for Support Vector Machines

9.8 Graphing Rational Functions

A MULTI-LEVEL IMAGE DESCRIPTION MODEL SCHEME BASED ON DIGITAL TOPOLOGY

SIMULATION OPTIMIZER AND OPTIMIZATION METHODS TESTING ON DISCRETE EVENT SIMULATIONS MODELS AND TESTING FUNCTIONS

ScienceDirect. Testing Optimization Methods on Discrete Event Simulation Models and Testing Functions

Optimization of fuzzy multi-company workers assignment problem with penalty using genetic algorithm

Research on Image Splicing Based on Weighted POISSON Fusion

An Improved Greedy Genetic Algorithm for Solving Travelling Salesman Problem

Himanshu Jain and Kalyanmoy Deb, Fellow, IEEE

Genetic Algorithms Coding Primer

MATRIX ALGORITHM OF SOLVING GRAPH CUTTING PROBLEM

Automated Planning for Feature Model Configuration based on Functional and Non-Functional Requirements

CS485/685 Computer Vision Spring 2012 Dr. George Bebis Programming Assignment 2 Due Date: 3/27/2012

Classifier Evasion: Models and Open Problems

Generell Topologi. Richard Williamson. May 6, 2013

A Novel Accurate Genetic Algorithm for Multivariable Systems

Global Caching, Inverse Roles and Fixpoint Logics

Nano Generalized Pre Homeomorphisms in Nano Topological Space

A Requirement Specification Language for Configuration Dynamics of Multiagent Systems

ES 240: Scientific and Engineering Computation. a function f(x) that can be written as a finite series of power functions like

Study and Analysis of Edge Detection and Implementation of Fuzzy Set. Theory Based Edge Detection Technique in Digital Images

ITU - Telecommunication Standardization Sector. G.fast: Far-end crosstalk in twisted pair cabling; measurements and modelling ABSTRACT

Fractal Art: Fractal Image and Music Generator

Lagrangian relaxations for multiple network alignment

Scalable Test Problems for Evolutionary Multi-Objective Optimization

Global Constraints. Combinatorial Problem Solving (CPS) Enric Rodríguez-Carbonell (based on materials by Javier Larrosa) February 22, 2019

EFFICIENT ATTRIBUTE REDUCTION ALGORITHM

Solution of the Hyperbolic Partial Differential Equation on Graphs and Digital Spaces: A Klein Bottle a Projective Plane and a 4D Sphere

Performance Assessment of the Hybrid Archive-based Micro Genetic Algorithm (AMGA) on the CEC09 Test Problems

A Classification System and Analysis for Aspect-Oriented Programs

Binary recursion. Unate functions. If a cover C(f) is unate in xj, x, then f is unate in xj. x

AUTOMATING THE DESIGN OF SOUND SYNTHESIS TECHNIQUES USING EVOLUTIONARY METHODS. Ricardo A. Garcia *

Collaborative Rough Clustering

/$ IEEE

Regenerating Codes for Errors and Erasures in Distributed Storage

ScienceDirect. Using Search Algorithms for Modeling Economic Processes

Optics Quiz #2 April 30, 2014

ROBUST FACE DETECTION UNDER CHALLENGES OF ROTATION, POSE AND OCCLUSION

The Graph of an Equation Graph the following by using a table of values and plotting points.

A SAR IMAGE REGISTRATION METHOD BASED ON SIFT ALGORITHM

MAPI Computer Vision. Multiple View Geometry

A Fault Model Centered Modeling Framework for Self-healing Computing Systems

A novel algorithm for multi-node load forecasting based on big data of distribution network

Scheduling in Multihop Wireless Networks without Back-pressure

AN 608: HST Jitter and BER Estimator Tool for Stratix IV GX and GT Devices

Classifier Evasion: Models and Open Problems

Reflection and Refraction

Bi- and Multi Level Game Theoretic Approaches in Mechanical Design

Xavier: A Robot Navigation Architecture Based on Partially Observable Markov Decision Process Models

1-2. Composition of Functions. OBJECTIVES Perform operations with functions. Find composite functions. Iterate functions using real numbers.

Walheer Barnabé. Topics in Mathematics Practical Session 2 - Topology & Convex

ON DECOMPOSITION OF FUZZY BԐ OPEN SETS

Design and Realization of user Behaviors Recommendation System Based on Association rules under Cloud Environment

Wave load formulae for prediction of wave-induced forces on a slender pile within pile groups

Compressed Sensing Image Reconstruction Based on Discrete Shearlet Transform

Binary Morphological Model in Refining Local Fitting Active Contour in Segmenting Weak/Missing Edges

arxiv:cs/ v1 [cs.lo] 4 Jul 2002

Using a Projected Subgradient Method to Solve a Constrained Optimization Problem for Separating an Arbitrary Set of Points into Uniform Segments

3-D TERRAIN RECONSTRUCTION WITH AERIAL PHOTOGRAPHY

Weighted Geodetic Convex Sets in A Graph

CS 161: Design and Analysis of Algorithms

Geometric Separators and the Parabolic Lift

Learning Implied Global Constraints

Keysight Technologies Specifying Calibration Standards and Kits for Keysight Vector Network Analyzers. Application Note

A Review of Evaluation of Optimal Binarization Technique for Character Segmentation in Historical Manuscripts

Comparison of Two Interactive Search Refinement Techniques

SUFFUSE: Simultaneous Fuzzy-Rough FeatureSample Selection

5.2 Properties of Rational functions

Research Article Synthesis of Test Scenarios Using UML Sequence Diagrams

An Efficient Configuration Methodology for Time-Division Multiplexed Single Resources

Proportional-Share Scheduling for Distributed Storage Systems

foldr CS 5010 Program Design Paradigms Lesson 5.4

Metamodeling for Multimodal Selection Functions in Evolutionary Multi-Objective Optimization

A Method of Sign Language Gesture Recognition Based on Contour Feature

Joint Congestion Control and Scheduling in Wireless Networks with Network Coding

Pre-defined class JFrame. Object & Class an analogy

Formalizing Cardinality-based Feature Models and their Staged Configuration

ROUGH SETS THEORY AND UNCERTAINTY INTO INFORMATION SYSTEM

ISSN: [Keswani* et al., 7(1): January, 2018] Impact Factor: 4.116

Structural Gate Decomposition for Depth- Optimal Technology Mapping in LUT- Based FPGA Designs

Genetic algorithms for job shop scheduling problems with alternative routings

On Generalizing Rough Set Theory

Multi-Objective Evolutionary Algorithms

An Automatic Sequential Smoothing Method for Processing Biomechanical Kinematic Signals

Literature Review On Implementing Binary Knapsack problem

Reducing the Bandwidth of a Sparse Matrix with Tabu Search

Open and Closed Sets

Research on Applications of Data Mining in Electronic Commerce. Xiuping YANG 1, a

Transcription:

Int J Contemp Math Sciences, Vol 6, 0, no 0, 45 465 A Proposed Approach or Solving Rough Bi-Level Programming Problems by Genetic Algorithm M S Osman Department o Basic Science, Higher Technological Institute Ramadan 0 th city, Egypt W Abd El-Wahed Department o Operations Researches and Decision Making aculty o Computers and Inormation, El-Menouia University Shebin El-Kom, Egypt Mervat M K El Shaei Department o Mathematics, aculty o Science Helwan University, Cairo, Egypt mervat_elshaei@yahoocom Hanaa B Abd El Wahab Department o Basic Science, Special Studies Academy, Cairo, Egypt Abstract We present a new ramework to hybridize the rough set theory with the bi-level programming problem, called Rough Bi-level Programming Problems (RBLPPsThis paper studies and designs a genetic algorithm (GA or solving (RBLPPs by constructing the itness unction o the upper level programming problems based on the deinition o through easible degree inally, a numerical example will be introduced to show the proposed method Keywords: Rough set, Rough Bi- level linear programming problems, Rough optimality, Rough easibility, Rough easible degree; itness unction; genetic algorithm

454 M S Osman et al Introduction Since it was pioneered by Pawalk in mid 980 s and [4] rough set theory has become a hot topic o great interest to researchers in several ields and has been applied to many domains such as pattern recognition, data mining, artiicial intelligence, image processing,machine learning, and medical application This new approach proved to be useul in many applications such as optimization theory or mathematical programming problems (MPPs in the crisp orm, the aim is to maximize or minimize an objective unction over certain set o easible solutions But in many practical situations, the decision maker may not be in a position to speciy the objective and/or the easible set precisely but rather can speciy them in a rough sense In such situations, it is desirable to use some rough programming type o modeling so as to provide more lexibility to the decision maker Towards this objective, we present a new ramework to hybridize the rough set theory with the bi-level programming problem, called Rough Bilevel Programming Problems (RBLPPs Since the roughness may appear in a Bi-level Programming Problems in many ways (eg the easible set may be rough and/or the goals may be rough, the deinition o rough Bi-level Programming Problems is not unique This leads us to propose a new classiication and characterization o the rough Bi-level Programming Problems (RBLPPs, according to the place o roughness in the problem We classiied the RBLPPs into the ollowing classes: Bi-level programming problems with rough easible set, crisp objective unction o the upper level and crisp objective unction o the lower Bi-level programming problems with rough easible set, crisp objective unction o the upper level and rough objective unction o the lower Bi-level programming problems with rough easible set, rough objective unction o the upper level and crisp objective unction o the lower 4 Bi-level programming problems with crisp easible set and rough objective unction o the upper level and crisp objective unction o the lower 5 Bi-level programming problems with crisp easible set and crisp objective unction o the upper level and rough objective unction o the lower 6 Bi-level programming problems with crisp easible set and rough objective unction o the upper level and rough objective unction o the lower 7 Bi-level programming problems with rough easible set and rough objective unction o the upper level and rough objective unction o the lower New deinitions concerning rough optimal sets, rough optimal value, rough global optimality and rough easibility were also proposed and discussed This paper studies and designs a genetic algorithm (GA o (RBLPPs by constructing the itness unction o the upper level programming problems based on the deinition o the rough easible degree

Rough bi-level programming problems 455 Rough set and approximation space Rough set theory has been proven to be an excellent mathematical tool dealing with vague description o objects [4] A undamental assumption in rough set theory is that any object rom a universe is perceived trough available inormation, and such inormation may not be suicient to characterize the object exactly Pawlak has proposed rough set methodology as a new approach in handling classiicatory analysis o vague concepts [4] In this methodology any vague concept is characterized by a pair o precise concepts called the lower and the upper approximations Rough set theory is based on equivalence relations describing partitions made o classes o indiscernible objects Let U be a non-empty inite set o objects, called the universe, and E U U be an equivalence relation on U The ordered pair A ( U, E is called an approximation space generated by E on U The equivalence relation E generates a partition U / E { Y, Y,, Y m where Y, Y,, Ym are the equivalence classes (also called elementary sets or granules generated by E, represent elementary portion o knowledge we are able to perceive due to E o the approximation space A In rough set theory, any subset M U is described by the elementary sets o A, and the two sets E ( M U { Y i U / E Y i M E ( M U { Y U / E Y I M φ i i are called the lower and the upper approximations o M, respectively Thereore, E ( M M E ( M The dierence between the upper and the lower approximations is called the boundary o M and is denoted by BN E (M E ( M E ( M The set M is called exact in A i BN E (M φ ; otherwise the set M is inexact (rough in A [5, 6, 8, 9] As we can see rom the deinition approximations are expressed in terms o granules o knowledge The lower approximation o a set is union o all granules which are entirely included in the set; the upper approximation - the union o all granules which have non empty intersection with the set; the boundary region o the set is the dierence between the upper and lower approximation [4] This deinition is clearly depicted in igure Classes o Rough Bi-level Programming Problems (RBLPPs be stated as: ( where x solve The most typical Bi-level Programming Problems (BLPPs can max ( x, x x

456 M S Osman et al max x ( x x, subject to x ( x, x M where and are called the objective unctions o the upper and lower level DM, and M is called the easible set o the problem In the above ormulation, it is assumed that all entries o, and M are deined in the crisp sense, and max is a strict imperative However, in many practical situations it may not be reasonable to require that the easible set or the objective unctions in bi-level programming problems be speciied in precise crisp terms In such situations, it is desirable to use some type o rough modeling and this leads to the concept o rough bi-level programming problems When decision is to be made in a rough environment, many possible modiications o the above bi-level programming model exist Thus, rough bi-level programming models are not uniquely deined as it will very much depend upon the type o roughness and its speciication as prescribed by the decision maker Thereore, the rough bi-level programming problems can be broadly classiied as: st Class: Bi-level programming problems with rough easible set, crisp objective unction o the upper level and crisp objective unction o the lower nd Class: Bi-level programming problems with rough easible set, crisp objective unction o the upper level and rough objective unction o the lower rd Class: Bi-level programming problems with rough easible set, rough objective unction o the upper level and crisp objective unction o the lower th 4 Class: Bi-level programming problems with crisp easible set and rough objective unction o the upper level and crisp objective unction o the lower th 5 Class: Bi-level programming problems with crisp easible set and crisp objective unction o the upper level and rough objective unction o the lower th 6 Class: Bi-level programming problems with crisp easible set and rough objective unction o the upper level and rough objective unction o the lower th 7 Class: Bi-level programming problems with rough easible set and rough objective unction o the upper level and rough objective unction o the lower In RBLPPs, wherever roughness exists, new concepts like rough easibility and rough global optimality come in the ront o our interest The rough easibility arises only in the st, nd, rd and 7 th classes, where solutions have dierent degrees o easibility (surely-easible, possible-easible, and surely-not easible On the other hand, the rough global optimality arises in all classes o the RBLPPs where solutions have dierent degrees o global optimality (surelyglobal optimal, possible- global optimal, and surely-not global optimal As a result o these new concepts, the optimal value o the objectives and the optimal set o the problem are deined in rough sense (

Rough bi-level programming problems 457 Deinition : In RBLPPs, the optimal value o the objective unction o the upper level DM is a rough real number, that is determined roughly by lower and upper bounds denoted by,, respectively Remark : I then the optimal value is exact, otherwise is rough Also, the single optimal set o the crisp bi-level programming problem is replaced by our optimal sets covering all possible degrees o easibility and optimality See table, igure Deinition : The set o all surely-easible, surely- global optimal solutions is denoted by O Deinition : The set o all surely-easible, possibly-global optimal solutions is denoted by O sp ss

458 M S Osman et al Deinition 4: The set o all possibly -easible, surely -global optimal solutions is denoted by O ps Optimality Possibly Surly O pp O Possibly ps O Surly sp O ss easibi lity Table : Possible degrees o easibility and optimality or problem (RBLPPs Deinition 5: The set o all possibly -easible, possibly -global optimal solutions is denoted by O pp 4 st Class RBLPPs Suppose that A ( U, E is an approximation space generated by equivalence relation E on an universe U A rough bi-level programming problem o the st class takes the ollowing orm: max ( x, x ( x where x solve max ( x, x (4 x subject to E ( M M E ( M where M U is a rough set in the approximation space A ( U, E representing the easible set o the problem The sets E ( M M and E ( M M represent the notion o rough easibility o problem ( - (4, where M is called the set o all possibly easible solutions and M is called the set o all surely-easible solutions On the other hand U M is called the set o all surely-not easible solutions [] Proposition : In problem ( - (4, the lower and upper bounds o the optimal objective value or the upper level problem are given by sup{ a, b sup{ a, c where a max ( x, x ( x x, M

Rough bi-level programming problems 459 b sup { min ( x, x c ( x x U Y U / E ( x, x Y Y M BN max ( x, x, M BN Deinition 6: A solution ( x,x M is surely -global optimal solution i ( x, x Deinition 7: A solution ( x,x M is possibly -global optimal solution i ( x, x Deinition 8: A solution ( x,x M is surely -not global optimal solution i ( x, x < Deinition 9: The optimal sets o the st class RBLPP or the upper level problem are deined as: O ss x, x M ( x, x O sp { ( { ( x, x M ( x, x { ( x, x M ( x, x ( x, x M ( x, x O ps O pp { Proposition : The lower and upper bounds o the optimal objective value or the lower level problem are given by sup{ a, b sup{ a, c where a max ( x, x b ( x, x M sup c U Y U / E ( x, x Y Y M BN max ( x, x ( x, x M BN min ( x, x Deinition 0: A solution ( x, x M is surely -global optimal solution i ( x, x Deinition : A solution ( x, x M is possibly -global optimal solution i ( x, x Deinition : A solution ( x, x M is surely not -global optimal solution i ( x, x < Deinition : The optimal sets o the st class RBLPP or the lower level problem are deined as: { ( x, x M ( x, x O ss

460 M S Osman et al O sp { ( x, x M ( x, x ( x, x M ( x, x ( x, x M ( x, x { { O ps O pp 5 Design o GA or RBLPPs The basic idea or solving RBLPP by GA is: irstly, choose the initial population satisying the constraints (by using the above deinitions and proposition,, the lower-level decision maker makes the corresponding rough optimal reaction and evaluate the individuals according to the itness unction constructed by the rough easible degree, until the optimal solution searched by the genetic operation over and over 6 Design o the itness unction To solve the problem RBLPPs by GA, the deinition o the easible degree is irstly introduced and the itness unction is constructed to solve the problem by GA Let d denote the large enough penalty interval o the rough easible region or each ( x, x X Deinition 4: Let θ [0,] denotes the rough easible degree o satisying the rough easible region, and describe it by the ollowing unction:, i x Y ( x 0 x Y ( x θ, i 0 < x Y ( x d (5 d 0, i x Y ( x > d where denotes the norm and Y ( x denote the rough optimal solution o lower level problem urthers, the itness unction o the GA can be stated as: eval ( vk ( ( x, x min θ where min is the rough minimal value o ( x, x on X

Rough bi-level programming problems 46 7 The algorithm Step Initialization, give the population scale M, the maximal iteration generation MAXGEN, and let the generation t 0 [7] Step The initial population, M individuals are randomly generated in X, making up the nitial population Step Evaluation the rough minimal value o upper level problem ( min and the rough optimal solution o the lower level ( Y ( x by using proposition, respectively Step 4 Computation o the itness unction Evaluate the itness value o the population according to ormula (5 Step 5 Selection, Select the individual by roulette wheel selection operator [] Step 6 Crossover, in this step, irst, a random number P c [0,] is generated This number is the percentage o the population on which the crossover is perormed Then, two individuals are selected randomly rom the population as parents Children are generated using the ollowing procedure: Random integer c is generated in the interval [, l ], where l is the number o components o an individual The c irst components o the children are the same components as respective parents (ie the irst child rom the irst parent and the second child rom the second parent The remaining components are selected according to the ollowing rules: (i The ( c + i th component o the irst child is replaced by the ( l i + th component o the second parent (ori,,, l c (ii The ( c + i th component o the second child is replaced by the ( l i + th component o the irst parent (ori,,, l c or example we assume c 5, we obtained the ollowing children Parents children 00 00 00 000 00 000 00 00 Note that the proposed operator generates individuals with more variety in comparison with the standard operator, because this operator can generate dierent children rom similar parents, where standard operators cannot [] Step 7 Mutation, In this step, irst, a random number Pm [0,] is generated This number is the percentage o the population on which the mutation perormed Then one individual is selected randomly rom the population An integer random number u is generated in the interval [, l ], where l is the length o the individual or generating the new individual, the u th component is changed to 0, i it was initially and to i it was initially 0 [] Step 8 Terminations, Jude the condition o the termination When t is larger than the maximal iteration number, stop the GA and output the rough optimal solution Otherwise, let t t +, turn to Step

46 M S Osman et al 8 An Example Let U be a universal set deined as x ( x, x { R x + 9 U x and let K be a polytope generated by the ollowing closed halplanes h x + x 0, h x x 0 h x x + 0, h4 x + x + 0 Suppose that E is an equivalence relation on U such that U / E E, E E {, { x U : x is an interior point o polytope K { x U : x is a boundary point o polytope K { x U : x is an exterior point o polytope K E E E Consider the ollowing st Class RBLPPs max ( x, x x x x M where x solve max ( x, x x M ( 5 x + x subject to M E U E, M E U E U E where M is a rough easible region in the approximation space A ( U, E and M, M are the lower and the upper approximation o M ; respectively Also, the boundary region o M is given by M BN E The solution By using the above genetic algorithm we ound the ollowing results: Give the population scale M 00, the maximal iteration generation MAXGEN 0, and let the generation t 0 Computation o the itness unction Evaluate the itness value o the population according to ormula (5 as: eval( vk ( ( x 5 x 044 θ To evaluate min, we use proposition and the given deinitions inding the rough minimal value o the upper level problem min x x 5 ( x + x sup{ a, b a max ( x, x ( 0, 0 044 Where, ( Q ( x, x in ( x, x M b sup ( x, x (0000, 0 65 E U { min ( x, x Y U / E, ( x, x Y Y M 044 BN min ( x, x (0000, 0 65 ( x x, E (6 (7

Rough bi-level programming problems 46 Q sup{ a, c Q sup ( x, x ( x, x E c max ( 0000, 0 65 ( x, x max ( x, x M ( x, x BN E ( x, x ( 0000, 0 6 5 044 min 0 44 inding the rough optimal value o the upper level problem : sup{ a, b a max x, x (,0 05 Q ( x, x in ( x, x M b sup ( ( x, x (,0 0 5 E U { min ( x, x Y U / E, ( x, x Y Y M BN 05 Q sup{ a, c Q sup x, x ( x, x E c max ( x, x M ( BN (5, 0 0 ( x, x max ( x, x E min ( x x E, ( x, x (,0 0 5 ( x, x (5, 0 0 0 [ 05, 0] inding the rough optimal sets or the upper level problem: O ss { ( x, x E U E : ( x, x 0 φ O sp { ( x, x E U E : ( x, x 0 5 {(,0 O ps { ( x, x E U E U E : ( x, x 0 {( 5,0 O PP {( x, x E U E U E : ( x, x 05 {( x, x : ( x 5 + x 05 But or evaluateθ, we ind Y ( x by using proposition such that: inding the rough optimal value o the lower level DM Y ( x x + or ixed x sup{ a, b a max ( x, x (, 4 ( x, x [,] Q in ( x, x (, ( x, x E b sup ( U min ( x, x min ( x, x (, Y U / E, x, x Y ( x, x E 4 Y M BN

464 M S Osman et al sup{ a, c Q sup ( x, x E ( x, x (, 5 c max ( x, x max ( x, x M ( x, x 5 [ 4, 5 ] BN E ( x, x (, 5 inding the rough optimal sets or the lower level O ss x, x E U E : x, x 5 { ( ( φ { ( x, x E E : ( x, x 4 { { ( x, x E U E E : ( x, x 5 { { ( x x E U E U E : ( x, 4 [,] O sp U O ps U O x PP, then Y ( x 4 and d 8 The operators o the genetic algorithm [, 7] are applied such that P 0 7, P 0 c m Obtain the optimal solution ( x, x, θ ( 5,, 055 Elapsed time is 445 minutes 9 Conclusions This paper proposes a new ormulation, classiication and deinition o the rough bi-level programming problems Only the st class o the RBLPPs is deined and its optimal sets are characterized in this work Also, it designs GA or solving RBLPPs o which the rough optimal solution o the lower-level problem is dependent on the upper-level problem Reerences [] SR Hejazi, A Memariani, G Jahanshahloo, M M Sepehri, Linear Bi-Level Programming Solution by Genetic Algorithm, Computers and Operations Research, 9 (00, 9-95 [] Z Michalewicz,Genetic Algorithms +Data StructuresEvolution Programs, Springer, New York, 99 [] MS Osman, E Lashein, EA Youness, TEM Atteya, Rough Mathematical Programming, Optimization A Journal o Mathematical Programming and Operations Research, 58 (009,-8

Rough bi-level programming problems 465 [4] Z Pawlak, Rough Sets, International Journal o Computer and Inormation Sciences, (98, 4-56 [5] H K Tripathy, B K Tripathy and P K Das, An intelligent Approach o Rough Set in Knowledge Discovery Databases, International Journal o Computer Science and Engineering, (007, 45-48 [6] S A Vinterbo, uzzy and Rough Sets, Decision Systems Group Brigham and Women s Hospital, Harvard Medical School, Spring, 00 [7] G Wang, Z Wan and X Wang, Solving Method or a Class o Bi-level Linear Programming based on Genetic Algorithms, School o Mathematics and Static, Wuhan University report China, (00,-7 [8] Y Y Yao, A Comparative Study o uzzy Sets and Rough Sets, International Journal o Inormation Sciences, 09 (004, 7-4 [9] H Zhang, H Liang, D Liu, Two New Operators in Rough Set Theory with Applications to uzzy Sets, International Journal o Inormation Sciences, 66 (004, 47-65 Received: August, 00