MULTIPLE CRITERIA DECISION MAKING : NETWORK FLOWS AND ITS FRAMEWORKS FOR APPLICATION IN HIGHER EDUCATION ADMINISTRATION

Abstract

The main purpose of this research is to promote further applications of Multiple Criteria Decision Making (MCDM) methods using existing techniques and tools. The Multiple Objective Network Flow (MONF) problem has been chosen as the methodological area, and higher education administration as the functional/ application area. Two separate literature reviews were first done on the indicated methodological and functional areas. From the reviews, a number of subareas where work is lacking have been identified. For the MONF problem, it was found that there has not been any reported work on the Tricriteria Network Flow (TCNF) problem, Tricriteria Integer Network Flow (TCINF) problem and Multiple Objective Integer Network Flow (MOINF) problem which utilizes the special properties of network flows. In addition to this, there has been only two reported uses of the interactive approach of handling multiple objectives and both are for the general MONF problem. Similarly, from the review of the reported applications of MCDM methods in higher education administration, the literature was lacking in the areas of scheduling (particularly timetabling) and in the use of MCDM models at the federal and state administration levels. The research then proceeded into rectifying some of the situations that have been identified as needing more work. For the MONF problem, the research focused on the Bicriteria Network Flow (BCNF), Bicriteria Integer Network Flow (BCINF), TCNF and TCINF. Some interesting characteristics of the efficient solutions of the BCNF and TCNF problems are first described. These characteristics are as observed from the use of a readily available user-friendly software package called DINAS (Dynamic Interactive Network Analysis System). Basically, these characteristics are associated with the special relationship that exists between efficient edges and cycles. By taking advantage of these special characteristics, the methods of integer adjusting the noninteger compromise solutions to the BCNF and TCNF problems obtained via DINAS are then proposed. An interesting aspect about these methods is that they only require the observation and manipulation of the decision values of the compromise solutions. With this adjustment, DINAS can now be used as a solution tool for BCINF and TCINF problems. With suitable solution tools for BCINF and TCINF problems, the research then focused on higher education administrations' effort to identify decision making situations that could appropriately be modelled as BCINF and TCINF problems. Timetabling was identified as an appropriate situation and, a BCINF based faculty course assignment model was proposed. The two objectives initially considered in the model are faculty course preferences and faculty course effectiveness. Later, in taking advantage of DINAS, another objective, i.e., faculty semester preferences, was introduced. Unlike the first two objectives, which are arc related, the third objective is node related. Fitting the BCINF-based faculty course assignment model into DINAS requires some adjustment as DINAS only works with zero-lower bound arcs while the BCINF-based model contains arcs with nonzero lower bounds. The required adjustment is also described in this research. Finally, suggestions for the possible expansion of the BCINF-based faculty course assignment model are given. The expansion is the addition of another criteria which turns the model into a TCINF-based model. The additional criteria is the efficiency rating of faculty member in other activities besides teaching, namely, research, supervision, administration, consultancy and community service. The efficiency rating is determined by a DEA (Data Envelopment Analysis) - based faculty evaluation model integrated with AHP (Analytic Hierarchy Process) for restricting the weights of output and input measures. The faculty evaluation model and the faculty courses assignment model together, serve as a framework towards the development of a systematic means of managing academic staff time between various activities. The research reported m this thesis contributes to the state-of-the-art understanding of MCDM in a number of ways. On the theoretical front, it describes some interesting characteristics of the efficient solutions of the BCNF and TCNF problems. The description then leads to contributions in the methodological aspect with the introduction of methods of integer adjusting the noninteger compromise solutions to the BCNF and TCNF problems. Finally, with suitable interactive solution tools for the BCINF and TCINF problems, contributions then focused on applications by suggesting some suitable frameworks for the use of BCINF and TCINF models in higher education administration.