MIS(MBA) - 信息系统协会中国分会.pptVIP

Sequencing of surgical operations 1||?wjCj Determination of weights: AHP (Analytical Hierarchy Process) Cost or importance of resources: grades of resource WSPT: optimal solution ? sequence of surgical operations Preference of doctors: with adjustment Least variance operation first Aseptic operation first (than non-aseptic) High-priority operation first Operations with general anesthesia first (than local anesthesia) Consecutive operations by the same surgeon * Appointment scheduling N: the set of surgical operations to be scheduled in a given session (normally a day), n = |N| Each surgical operations i?N : a random processing time pi(?), ??? ?: the set of all outcomes Integer valued and bounded Decision variable: A = (A1, A2, …, An, An+1=?) Si(?) and Ci(?): the start time and completion time of i?N in scenario ???, with C(?)=0 Si(?) = max(Ci-1(?), Ai) Ci(?) = Si(?)+ pi(?) * Cost functions Overage cost oi(·): R+?R, cost of customer waiting Underage cost ui(·): R+?R, cost of doctor idling Convex (piecewise linear) and non-decreasing ui(0) = oi(0)=0 Formulation of the problem MinA E?{{?ni=1[oi((Ci(?)-Ai+1)+)+ui((Ai+1-Ci(?))+)]}, s.t. Ci(?)=max{Ci-1(?), Ai}+pi(?), ?i?N, ????} Linear cost function case F(A) = E?{?ni=1[oi*(Ci(?)-Ai+1)++ui*(Ai+1-Ci(?))+)]} * Equivalent to problem (P): MinA,CE?{?ni=1[oi((Ci(?)-Ai+1)+)+ui((Ai+1-Ci(?))+)]} s.t. Ci(?)=max{Ci-1(?), Ai}+pi(?), ?i?N, ???? Theorem 2. If both oi( ) and ui( ) are piecewise linear and the break points are integers, then (P) has an integer solution Convex relaxation of (P) MinA G(A) := E?[F?(A)], where: F?(A) := minC h?(C, A) s.t. Ci? Ai+pi(?), ?i=2, …, n+1 Ci? Ci-1+pi(?), ?i=2, …, n+1 C1= p1(?) * Theorem 3. If h?(C, A) is non-decreasing in (C2, …, Cn) for each ??? and A, then the convex relaxation is exact. Convex program: polynomial-time solvable O(?(n)EOn2 logpmax) Preference of patients Hospitals offer a set of time slots Patients request a a set of time slots Availability of patient