Cheng, SiXin (1999) Fuzzy clustering with an application to scheduling. Masters thesis, Concordia University.
Usually, the generation of an optimal schedule is a costly and time-consuming process. This process requires expensive computational software and hardware. Scheduling problem modeling using human expert knowledge is promising and flexible in dealing with real world applications. Unfortunately, human expert knowledge may not be available in all cases, and human experts may not be able to explain their knowledge explicitly. A new scheduling decision learning approach is introduced in this thesis. A subtractive clustering based system identification method is developed to learn the scheduling decision mechanism from an existing schedule. It is utilized to build a fuzzy expert model. The existing schedule can be an optimal schedule developed using an optimization method or a schedule generated by a human expert. The fuzzy expert model is then used to generate new schedules for other problems following the decision mechanism it learned. The implementation of this method is demonstrated by modeling a single machine weighted flowtime problem. Furthermore, selective subtractive clustering and modified subtractive clustering algorithms are developed and used to improve knowledge extraction. Those algorithms can also be used to model nonlinear and spiral systems using the clustering based system identification, such as function approximation applications and pattern classification applications when the information about the system is scarce.
|Divisions:||Concordia University > Faculty of Engineering and Computer Science > Mechanical and Industrial Engineering|
|Item Type:||Thesis (Masters)|
|Pagination:||xv, 117 leaves : ill. ; 29 cm.|
|Degree Name:||Theses (M.A.Sc.)|
|Program:||Mechanical and Industrial Engineering|
|Thesis Supervisor(s):||Demirli, K.|
|Deposited By:||Concordia University Libraries|
|Deposited On:||27 Aug 2009 17:13|
|Last Modified:||08 Dec 2010 15:16|
Repository Staff Only: item control page
Downloads per month over past year