DECISION SUPPORT SYSTEM OR MANAGING EDUCATIONAL CAPACITY UTILIZATION IN UNIVERSITIES Svetlana Vinnik 1, Marc H. Scholl 2 Abstract Decision-making in the fiel of acaemic planning involves extensive analysis of large ata volumes originating from multiple systems. Acaemic workloa management is concerne with istributing teaching resources in orer to aequately support the university s eucational framework (faculties, egrees, courses, amission policies, teaching workloa etc). We propose a methoology for assessing eucational capacity an planning its istribution an utilization in universities. The approach has been implemente as a ecision support system allowing simulation an evaluation of various proposals an scenarios. The system integrates the input ata from relevant sources into an autonomous ata warehouse, while the client front-en ensures aequate output presentation to users so as reveal significant etails an epenencies. Applying the system as on-the-fly ecision-support facility by the policy-makers has resulte in significant acceleration of planning proceures, raise the overall awareness with respect to the unerlying methoology an, ultimately, enable more efficient acaemic aministration. Inex Terms acaemic Decision Support System, eucational capacity utilization, teaching workloa management, acaemic resource planning. INTRODUCTION Acaemic resource planning is a highly complex aministrative proceure base on extensive analysis of the entire ata relate to the eucational framework, such as teaching resources, offere egrees, course structure an curricula, stuent numbers etc. State-of-the-art ecisionmaking within most universities aroun the globe has the form of an argumentative pie-cutting barely backe up by any soli quantitative analysis. However, the emergence of avance information technologies has altere the operational environment of universities worl-wie offering them an opportunity to move on towars more systematic an efficient management of their assets. Accurate computational moel, comprehensible methoology, complete an consistent ata basis an a frienly output presentation are of paramount importance for avance ecision support. requently experience problems inclue unavailability of the ata in an appropriate form an lack of tools an approaches for its evaluation. rom the early ays of information systems aministrative acaemic processes such as effective resource istribution, teaching personnel management, automation of stuent amission an registration, stuent performance, retention an ismiss, to name the major ones, have been among the hottest eucationalist issues. irst attempts to implement simulation moels for hanling eucational resource management go back to the 60-ies [12] with renewe enthusiasm in the 90-ies apparently encourage by the overall avancement of information technology. In the 80-ies the acaemic ecision theory focuse mainly on formulating the general principles an approaches of the moel-base ecision support systems (DSS) for acaemic environments [10], [14]. Various acaemic DSS for resource allocation [1], [8], performance assessment [4], course scheuling [6], amission policy [7], avising [13], stuent profile evolution [2] an strategic planning [1] have previously been propose. More recent attempts circle aroun ata warehouse approaches which enable ata collection across ecentralize applications necessary for solving of complex aministrative problems [9]. The goal of our research is to contribute to the next generation of acaemic DSS base on the ata warehouse technology with incorporate ata mining an knowlege iscovery functionalities. Decision-making is supporte primarily by means of intelligent information presentation an by proviing options for its explorative analysis. ocusing on the exploration rather than on generating reaymae solutions has the avantage of ensuring the moel s aaptability an applicability for a wie range of problems. Our DSS targets to support the aministrative task of planning the university s eucational capacity in terms of the number of stuents its courses can accommoate uner the specifie resource constraints. Decision-makers are able to evaluate various strategies an generate forecasts by means of simulating with the input ata. To keep the moel manageable an intuitive an to avoi functional explosion we opte for a single bottle-neck resource of the eucational capacity, which is the teaching staff. rom experience, staff availability is by far the most crucial resource constraint, expensive an harly ajustable in the short-term compare to other resources involve, such as facilities, buget, appliances, materials etc. 1 Svetlana Vinnik, University of Konstanz, Department of Comp. & Inf. Science, P.O.Box D 188, 78457 Konstanz, Germany, vinnik@inf.uni-konstanz.e 2 Marc H. Scholl, University of Konstanz, Department of Comp. & Inf. Science, P.O.Box D 188, 78457 Konstanz, Germany, scholl@inf.uni-konstanz.e
Our contribution is basically twofol: 1) to propose the overall methoology for eucational capacity analysis, an 2) to esign its aequate ata warehouse implementation. The paper is structure as follows: Section 2 introuces the university ata moel followe by the eucational capacity methoology in Section 3; in Section 4 we elaborate on the implementation issues; interactive an explorative aspects are iscusse in Section 5. We conclue by a summary of our contribution an proposals for future work. MODELING THE UNIVERSITY STRUCTURE Universities worl-wie ten to have a hierarchical structure consisting of faculties, egrees, an courses. aculties are the basic aministrative units, each responsible for normally a single scientific iscipline in terms of offering stuy programs an courses relate to it. Multiisciplinary faculties, in case their isciplines were groupe for merely aministrative reasons, can be further ivie into subfaculties, calle units, to process each iscipline separately. Provision of both the aministrative (faculties) an the scientific (units) structure enables istinction between interisciplinary an interfaculty relationships. Let us consier the acaemic processes in terms of eucational supply an eman relationships, with faculties as suppliers of eucational services an stuents as their consumers. Obviously, resource utilization is full an balance if per-faculty supply correspons to its eman. Since faculties/units are the main actors in the process of resource allocation, we procee by substantiating the concept of eucational supply an eman on a per-faculty basis. Most curricular activities (lectures, tutorials, exams), henceforth referre to as courses, span one semester or can be mappe accoringly, so that semester appears to be an appropriate timeframe unit for resource utilization analysis. Eucational Supply Eucational supply measures the available teaching capacities in terms of the volume of services the faculty s staff provies to stuents. Teaching resources are classifie into position groups (e.g., professor, research assistant, etc.) with a specific teaching loa assigne to each group by respective legislation. Teaching loa efines the number of acaemic hours per week, enote semester perios per week, or SPW, to be investe in teaching. Thereby, the potential supply results from the total amount of available SPW whereas its particular instance consists of the courses actually offere. igure 1 visualizes the above supply concept using the example of a two-isciplinary faculty. Eucational Deman Eucational eman escribes the consumption of the faculty s acaemic services by the stuents who atten courses accoring to their respective curricula. IGURE 1. EDUCATIONAL SUPPLY O A ACULTY. aculties are responsible for stuy programs in their respective scientific iscipline an as such supervise the stuents enrolle therein. Each stuy program is characterize by a subject an a egree. In case of composite egrees each major an minor subject has to be processe separately since they normally have ifferent supervising faculties. Yet another special case is an increasingly popular class of interisciplinary programs whose very subjects are composite (e.g., Bioinformatics). Such programs are thus supervise by more than one faculty. igure 2 shows the composition of the eman at the example of two faculties. IGURE 2. EDUCATIONAL DEMAND O A ACULTY. Cross-aculty Relationships So far both the supply an the eman have been presente as having prevailingly hierarchical structures
(most faculty sub-trees o not overlap). In case of a strict hierarchy each faculty woul manage exclusively its own available resources an supervise stuy programs. However, in reality there exist iverse cross-faculty epenencies on the part of both the supply an especially the eman. These intensive interrelations make isolate per-faculty optimization of capacity utilization unfeasible. Major interisciplinary issues an cross-faculty interactions arise at the following levels: Courses: some courses with complex subjects are offere as a joint effort of teaching staff belonging to ifferent faculties. Stuy programs: stuy programs having an interisciplinary subject are typically jointly supervise by multiple faculties responsible for respective subisciplines (as in igure 2). Degrees: Those egrees with subivision into major/minor subjects (e.g., teacher s eucation egree) frequently encourage combination of non-relate subjects. Stuents enrolle in such egrees are registere within multiple faculties. Curricula: Curricula of most stuy programs contain blocks of courses offere by non-supervising faculties, thus eclaring their epenency on importe services. Examples of the cross-faculty interactions liste above are presente in igure 3. Cross-faculty teaching services can be escribe as exports-imports relationships of the faculties. rom the point of view of a single faculty, awareness of its expecte exports volume is inispensable in orer to account for it when estimating the capacity available for servicing its own supervise programs. The next step is to provie a methoology for measuring eucational supply an eman. IGURE 3. MAPPING CROSS-ACULTY DEPENDENCIES (SIMPLIIED RAGMENT). METHODOLOGY As presente is Section 2, the acaemic capacity utilization is all about proviing an consuming curricular activities. The overall eucational capacity is the total of the capacities of all offere courses. Each course is characterize by the volume measure in SPW an the support relation upper-bouning the number of participants. Aitionally, course types can be weighte ifferently epening on the preparation-intensiveness on behalf of the teaching staff. Thereby, the teacher-hours-per-stuent cost of some course C of type T with support relation N, calle the course s curricular value (CV), can be estimate as follows: CV SPW weight C T C =. (1) NC Curricular value of some particular egree is thus the total of the CVs of all the courses specifie in the respective curriculum. It thus represents the necessary number of teacher perios per-stuent necessary for completing that egree. Table I shows an example of estimating the CV from the egree s curriculum. TABLE I CURRICULAR VALUE O A STUDY PROGRAM (RAGMENT) seme- Course title SWS support offere by ster relation faculty CV Introuction into CS I 6 150 Comp. Sc. 0.04 1 st Operating systems 4 50 Comp. Sc. 0.08 Maths for Programmer 6 100 Maths 0.06 Algorithms&Data Struct. 6 100 Comp. Sc. 0.06 2 n Statistical Methos 4 200 Statistics 0.02 Information Systems 6 100 Comp. Sc. 0.06 3 r Inform. Managemen 6 100 Comp. Sc. 0.06 Practical Assignment 4 20 Comp. Sc. 0.5 6 th Project Management 4 50 Economics 0.08 Patent Law 2 100 Law 0.02 Total Curricular Value (B.Sc. in Information Engineering) 3.85 Summing up the course CVs groupe by the servicing faculty (rows of the same color in Table I) in the curriculum of egree yiels the per-stuent eman within for each involve faculty. Such per-faculty portions in the total egree s CV are calle faculty s curricular contributions (CC) within a egree. Intuitively, a convenient an compact way of isplaying all faculty/egree interactions is by arranging the CCs into a matrix with faculties as columns an egrees (clustere by supervising faculty) as rows. Each cell [,] thereby escribes faculty s CC in egree, as shown in Table II. Deriving faculties CC from the curricula, however, may be rather aggravate if the latter are efine in a flexible manner, i.e. allowing stuents to iniviually select courses following some general selection rules. Our propose solution is to synthesize a representative curriculum by combining formal curriculum efinition with the analysis of the actual recent course attenance statistics. or a single semester, the per-stuent eman value in
any matrix cell [,], ivie by the length of egree in semesters an multiplie with number of enrollments in in that semester, escribes s eucational eman within faculty : CC eman = # enrollments. (2) # semesters Summarize vertically, those values prouce the total require resources of faculty. The ratio between the total eman towars an its available teaching resources in SPW shows the faculty s teaching capacity utilization: Maths Inf. Sc. Biology Economy interisc CapUtilization eman all =. (3) TeachingLoa TABLE II EXPORTS/IMPORTS RELATIONSHIPS O THE ACULTIES (RAGMENT) aculty/degree Maths Inf. Sc. Biology Economy TOTAL Maths, Dipl. 1.80 0.30 0.02 3.20 Maths, T.E., 1 1.10 0.05 2.45 Maths, T.E., 2 0.55 1.82 Inf. Eng., B.Sc. 0.04 2.33 0.03 2.60 Inf. Eng., M.Sc. 1.41 1.85 Comp. Sc., T.E., 2 0.04 1.80 2.25 Biology, Dipl. 0.30 5.11 6.44 Biology, B.Sc. 0.17 0.02 2.97 4.02 Biology, M.Sc. 1.85 2.25 Economics, Dipl. 0.92 0.02 2.43 3.40 Economics, T.E., 1 0.74 0.01 2.11 2.74 Economics, B.Sc. 0.80 0.04 2.56 3.05 Economics, M.Sc. 0.06 1.40 1.74 Inf. Manag., B.Sc. 1.72 2.12 2.84 Inf. Manag., M.Sc. 0.90 1.65 1.50 Life Science, B.Sc. 0.05 0.02 2.78 4.55 urther computational etails an moel parameters are omitte here ue to limite space. The introuce methoology can be applie for solving a wie range of problems relate to acaemic capacity planning, such as: etermining the faculty s amission capacity for a specifie egree setting (presente in [15]); locating the bottle-neck faculty when moifing the egree structure; computing the necessary ajustment of teaching resources for supporting a certain szenario; checking whether there are any iscrepancies between offere egrees an courses. Other tasks can be easily enable by efining the respective reporting structure (input, options, output). IMPLEMENTATION Our chosen solution is a atabase-enable webapplication since it best fulfils the requirements of a DSS with high availability an a ifferentiate multi-user access. The system runs completely server-sie (ynamic content is generate by PHP scripts, complex visualizations are implemente in Java) so that all a client nees to access it is just a web browser an a network connection. The major challenge is the pre-processing phase in which the entire input ata has to be ientifie, collecte an integrate into the ata warehouse. Depening on the types of the systems involve, the quality of the ata (e.g., consistency, completeness, an format) an the legal ata access constraints this process may take up to several months. ortunately, the moel can operate on incomplete ata for solving less complex tasks. In case of missing ata, the user is prompte to specify a let-out (e.g., to use efault values, generate assumptions or aggregate over historical ata) or even to manually fill the gaps in the require input. Expert knowlege of the users is generally a valuable asset an a significant contribution to the system s reliability since it is harly feasible to ensure complete an faultless automate ata extraction an analysis. Simulation moe is realize by means of creating a copy of the ata subset require for solving the specifie task an allowing the user to manipulate that input an generate the esire reporting ocuments. Both the unerlying ata an the reports of simulation scenarios can be store, reloae, re-processe an share among multiple users. SIMULATION AND VISUAL EXPLORATION The system s graphical user interface has been esigne with the objective to encourage both strictly guie as well as freewheeling user interaction moes. Most ecision processes require combination of both approaches anyway - the expert woul first use their intuition to explore potentially relevant ata an woul then apply particular analytical tools for assessing particular problem areas. Guie interaction moe is implemente as an analytical toolbox with output report generation for a pre-efine set of acaemic problems an contains the following steps: (1) select one of the preefine tasks from the list; (2) select the type of report to generate; (3) select the report options (level of etail, aggregations, assumptions, error reporting, input inconsistency hanling, etc.); (4) ajust efault values an assumptions; (5) specify the input ata requirements an options; (6) ajust the input ata to map the esire scenario; (7) run the scenario an generate the output report; (8) interactively explore the output (using zooming, aggregating, etails-on-eman an other techniques);
(9) repeat steps (6)-(8) to iterate to a satisfactory result or to collect the information require for ecision-making. We conclue our presentation by proucing some fragments from the system s report generate for stuying the effects of introucing a new interisciplinary egree on the eucational capacity utilization. The screenshots in igure 4 isplay the ata in a top-own fashion, starting from the whole university overview to a single egree. IGURE 4. EXPLORING THE RESULTS O A USER-DEINED SCENARIO (RAGMENTS): A) PER-ACULTY CAPACITY UTILIZATION RATIO B) EDUCATIONAL DEMAND DISTRIBUTION WITHIN A ACULTY C) CURRICULAR VALUE COMPOSITION O A DEGREE CONCLUSION We have focuse on the problem of offering reliable ecision support to the process of balancing eucational eman an supply in universities. We first introuce the concept of the acaemic structure an the methoology for assessing the eucational capacity base on correct an accurate mapping of cross-faculty epenencies. The moel has been implemente as a group DSS for online construction an evaluation of acaemic scenarios. The system integrates ata from heterogeneous university systems. Decision support functionality is realize by offering reporting tools for solving particular capacityrelate tasks as well as by allowing users to navigate through the ata, query it, generate interactive visualizations an explore those for retrieving interesting etails. Our future work will be irecte towars improving the ata integration routines an enhancing the user interface to enable intuitive an flexible interactive visual exploration of the accumulate ata with incorporate ata mining techniques for expert tren analysis. REERENCES [1] Barlas, Y. et al., "A Dynamic Simulation Game (UNIGAME) for Strategic University Management", Simulation Gaming, Vol.31, 2000, 331-358. [2] Boren V.M., Dalphin, J.., Carson, L., "Simulating the Effect of Stuent Profile Changes on Retention an Grauation Rates: A Markov Chain Analysis", 38th Annual orum of the Association for Institutional Research, Minneapolis, MN, 1998. [3] Casper, C.A., Henry, M.S, "Developing Performance-Oriente Moels for University Resource Allocation", Research in Higher Eucation, Vol.42, 2001, 353-376 [4] Deniz, D. Z., Ersan, I., "Using an Acaemic DSS for Stuent, Course an Program Assessment", Proceeings of the ICEE 2001 Conference, Oslo, Norway, 2001 [5] Deniz, D.Z., Uyguroglu, M., Yavuz, H., "Departmental Workloa Aministration Using Group orecasting in Universities", Int. Conf. on Engineering Eucation (ICEE'02), Manchester, UK, 2002. [6] Deris, S.B. et al., "University timetabling by constraint-base reasoning: A case stuy", Journal of the Operational Research Society, Vol.48, 1997, 1178-1190. [7] Elimam, A.A., "A Decision Support System for University Amission Policies", European Journal of Operational Research, Vol.50, No.2, 1991, 140-156. [8] ranz, L. S. et al., "An Aaptive Decision Support System for Acaemic Resource Planning", Decision Sciences, Vol.12, No.2, 1981, 276-293. [9] Ingham, J., "Data Warehousing: A Tool for the Outcomes Assessment Process", IEEE Trans. on Eucation, Vol.43, No.2, 2002, 132-136. [10] Kassicieh, S.K., Nowak J.W., "Decision Support Systems in Acaemic Planning: Important Consierations an Issues", Information Processing an Management, Vol.22, No.5, 1986, 395-403. [11] Kwak, N., Lee, C., "A Multicriteria Decision-Making Approach to University Resource Allocations an Information Infrastructure Planning", European Journal of Operational Research, 110/2, 1998, 234-242. [12] Levine, J.B., "Application of the CAMPUS simulation moels to the major planning ecisions of a large university", Proceeings of the secon conference on Applications of simulations, New York, Unite States, 1968, 317-328. [13] Murray, W.S., Le Blanc L., "A Decision Support System for Acaemic Avising", Proc. of 1995 ACM Symposium on Applie Computing, Nashville, TN, USA, 1995. [14] Turban, E., isher, J.C., Altman S., "Decision Support Systems in Acaemic Aministration", Journal of Eucational Aministration, Vol.26, No.1, 1988, 97-113. [15] Vinnik, S., Scholl, M.H., "UNICAP: Efficient Decision Support for Acaemic Resource an Capacity Management", Proceeings of the TED Conference on e-government (TCGOV2005), Bozen-Bolzano, Italy, 2005.