Available online at www.sciencedirect.com
Computers & Operations Research 31 (2004) 985 – 999
www.elsevier.com/locate/dsw
Object-oriented decision support system modelling for multicriteria decision making in natural resource management Dingfei Liu∗ , Theodor J Stewart Department of Statistical Sciences, University of Cape Town, Rondebosch 7701, South Africa
Abstract There is a great need for domain analysis and modelling for decision support systems (DSSs) with an e0ective methodology to assist the development of DSSs. This paper discusses the object-oriented modelling of DSSs for multicriteria decision making (MCDM) in natural resource management. This approach of DSS modelling is integrated into the uniform framework based on object orientation for both MCDM and DSS modelling. The general system requirements for DSSs for MCDM in natural resource management decision problems are captured. Primary classes, including decision elements and DSS components, for both MCDM decision analysis and DSS modelling are then extracted. General DSS system architecture is proposed. The system development of a DSS for water resources management based on the DSS model has demonstrated both e6ciency of the development process and e0ectiveness of the system developed. ? 2003 Elsevier Ltd. All rights reserved. Keywords: Object orientation; MCDM; DSS modelling; Natural resource management
1. Introduction Decision making for natural resource management is very complex, involving various stakeholders and communication mechanisms among them. Consensus is needed amongst the di0erent groups or individuals who have con;icting social, economic and political interests. In such a complex problem as natural resources planning, there are issues that cannot be adequately modelled because they are qualitative in nature, unknown, or un-revealed by the decision makers [1]. For the last few years, multicriteria decision making (MCDM) has been applied vigorously in the support of decision making in natural resource management. Indeed, the well-developed @eld ∗
Corresponding author. Department of Decision Sciences & MIS, Concordia University, Montreal, Canada H3G 1M8. Fax: +1-514-848 2824. E-mail address:
[email protected] (D. Liu). 0305-0548/$ - see front matter ? 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0305-0548(03)00047-9
986
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
of MCDM o0ers a number of quite divergent methodologies, including utility and value theory, goal programming and reference point approaches, outranking approaches, game theory, the analytic hierarchy process, etc. [2–6], to problems of natural resource planning [7,8]. Ellis [9] adds that the environmental operational research projects have succeeded thanks to multi-objective methodology since environmental problems are fundamentally multi-objective in nature. Hallefjord and Jornsten [10] emphasise that the purpose of applying MCDM to long-term planning problems is to explore solutions, to generate alternative strategies, and to gain insight into the problem, rather than to @nd optimum solutions. Complexity and long development times are inherent in building DSSs, limiting their wide use as a result [11]. Building a DSS requires signi@cant expertise both in decision analysis and in programming. The task is further complicated by the fact that a DSS may also need to connect in real time with other enterprise applications. These reasons are claimed [11] to be the deterrents to the wide use of DSSs. A DSS generator is suggested as a current solution for developing an application speci@c DSS. Using a DSS generator reduces DSS development to a decision analysis task, which requires expertise in decision analysis and mathematical modelling rather than programming skills. However, in [11], not much attention is paid to the di6culties for non-analysts to model the decision problem with expertise from decision analysis in such a speci@c DSS. In MCDM in natural resource management, the issues of complexity and di6culty of DSS development and utilisation for practical decision problems may be worse than for general DSSs because of the complex and multidisciplinary nature of MCDM natural resource problems. As discussed above, natural resource management is complex due to its multiple dynamic aspects, and the existence of di0erent role players in the natural resource allocation who have con;icting objectives. Natural resource DSSs are complex applications [12], possibly integrating di0erent advanced technologies and requiring high research and development e0orts. DSSs for natural resource management are time consuming and costly to develop and maintain. As a result, these DSSs are di6cult to adapt to rapidly changing decision environments. A sound DSS generator in the @eld of natural resource management is an invaluable asset to the practitioners of decision analysis and also to stakeholders in a decision problem. In the area of group decision support systems (GDSSs), despite the increasing number of published studies, little progress has been made toward theoretical models that integrate the existing empirical observations and that o0er guidelines in developing e0ective group systems [13]. The motivation for the theoretical models stems from two interrelated sources: the need to reconcile the inconsistent results across studies and an assumption that theory-based research will advance the area. The inconsistencies in empirical results are argued to be at least partly due to the lack of well-articulated theoretical models for developing hypotheses and interpreting results. A theoretical model can provide the commonality necessary amongst researchers to build a clear and rapid understanding of GDSSs. Wang [14] notices that there is no commonly used tool for the domain analysis and modelling of DSSs, despite the vast amount of literature on the research of DSS theory and applications. Little research has been forthcoming with regard to analysis techniques for DSSs. The objective of using a DSS domain analysis technique is to determine the requirements for resources in supporting certain types of decisions as well as the possible paths involved in such decisions. The outcome of a DSS model should be of assistance in the design and implementation of the DSS.
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
987
DSSs for MCDM in natural resource management are GDSSs since multiple decision participants are involved in the decision making process. Schmoldt and Peterson [15] suggest that group decision making with multiple values is becoming increasingly important in natural resource management and associated scienti@c applications. Multiple values are treated coincidentally in time and space, while multiple resource specialists and multiple stakeholders must be included in the decision processes. It seems that there is a great need for domain analysis and modelling for DSSs, especially for those that support MCDM in natural resource management, with an e0ective methodology to assist the development of DSSs as well as determining the resources and paths for decision making. Modelling can manage the complexity in decision problems and DSSs for natural resource management. It is futile to attempt to understand all the aspects of a decision problem and the decision making procedure and to implement an e6cient and e0ective DSS without some form of framework or models. This modelling methodology can model these DSSs in a scienti@c way and reduces the requirements for artistic skills in modelling DSSs. A DSS for a speci@c natural resource decision problem may be speci@ed as a sketch of the DSS model resulting from the modelling. The DSS model can deal with the complexity of DSS development and make the development of a speci@c DSS a0ordable in terms of both time and cost. Ideally this methodology should be integrated with that for modelling multicriteria decision problems since a uniform methodology and philosophy is very desirable in MCDM and DSS development to allow the improvement of e0ectiveness and e6ciency in both aspects [16]. The DSS model is the mechanism to bridge the gap between the DSS researchers and decision analysis practitioners as it renders the DSS design, implementation and evaluation understandable to decision practitioners and DSS researcher alike. Finally, an ideal DSS model can support the development of network-based GDSSs for all the phases of natural resource decision making. Object orientation [14,17–21] has demonstrated a promising future as a pragmatic methodology in the modelling of DSSs for MCDM in natural resource management. Due to its success in the development of general software systems, object orientation has been applied in some aspects of DSS development [20–31]. These application aspects mainly include programming, user interface development, architectural design, model management and database management. However, the literature is very limited on techniques for object-oriented DSS analysis and for object-oriented support of the DSS development life-cycle (including analysis, design, and implementation). Most importantly, little research has been conducted on the domain analysis and the object-oriented modelling of DSSs in general or of DSSs for MCDM in natural resource management in particular. Few comprehensive object-oriented methodological principles have been proposed for the incorporation of decision analysis and DSS modelling [16]. This paper discusses an object-oriented modelling of DSSs for MCDM in natural resource management. This approach of DSS modelling is integrated into the uniform framework based on object orientation for both MCDM (Multicriteria Decision Making) and DSS modelling [16]. The general system requirements and various classes are actually essential part of an object-oriented DSS model since they capture generic aspects of DSSs. The capture of DSS requirements is based on the understanding of the fundamental functions and the system environment of DSSs. Decision making processes are analysed and the system infrastructure is determined. In addition, DSS evaluation principles can be applied to the requirement analysis to ensure that these principles are met by the DSS under development.
988
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
Problem Environment
Facilitator
Criterion Hierarchy
Domain Expert
Stakeholders
Alternative Set
Choice Implementation
Fig. 1. General MCDM decision making procedure for natural decision problems.
The remainder of this paper is organised as follows. A conceptual framework based on object orientation is proposed for the general MCDM decision making procedure for natural resource management in Section 2. Section 3 @rst de@nes the system users along with the system context for DSSs. The main system functions of DSSs are captured and are represented with use cases. Classes of decision elements and other basic system components are identi@ed. Section 4 discusses DSS subsystems, system architecture, and the utilisation of the DSS model in the development of DSSs. Section 5 introduces the development of a DSS for water resources management based on the DSS model. Conclusions are contained in Section 6.
2. Modelling the MCDM decision making procedure A general framework of MCDM decision making concepts can help the understanding of the basic considerations for MCDM in natural resource management. A conceptual framework, which is shown in Fig. 1, is proposed for the general MCDM decision making procedure for natural resource management and similar strategic multi-stakeholder decision contexts. Fig. 1 shows the group structure of decision participants and the group interaction. According to Phillips [32] a group decision making procedure should be able to help structure the thinking of the group, and deal with group dynamics. Its method of operation should be understandable to participants and it should be ;exible. In the context of resource allocation problems with high levels of con;ict and antagonism, these features seem to be especially important. Di0erent viewpoints from various actors, including stakeholders, a decision facilitator/analyst and domain experts, which we shall term “decision participants”, can be observed, as well as an overall perspective of the decision making processes. Group decision participants communicate with each other under the guidance of a facilitator/decision analyst, who is skilled in the processes of group discussion and decision analysis, and has some understanding (at least basic knowledge) in the context of the problem at hand. The facilitator assists the participants in brainstorming the problem and structuring the decisions facing the group. The problem structure and value judgements come from the group.
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
989
Criteria are usually constructed hierarchically for MCDM problems. In natural resource management, con;icting criteria re;ect the lack of consensus amongst the di0erent groups or individuals who have incompatible social, economic and political interests, and who bene@t from the exploitation of the natural resources. Keeney and Rai0a [33], and Von Winterfeldt and Edwards [34] show that it is possible even in complex problems to structure management criteria hierarchically. At the top of the hierarchy is a general statement of the overall criteria (e.g., economy), which need to be progressively better de@ned at lower levels of the hierarchy to clarify the vagueness of the meanings of the broad criteria. This re@nement of criteria will continue until they become very speci@c (e.g., employment rate) at the lowest level. These lowest objectives need to be operationally meaningful in the sense that a decision alternative can be more or less unambiguously judged according to this attribute. The decision making procedure shown in Fig. 1 is iterative. Decision participants may @rst identify various alternatives as well as objectives (criteria) as a result of brainstorming. A choice is made out of the evaluation of alternatives according to the criterion hierarchy constructed by using MCDM techniques, such as those from multiattribute utility theory (MAUT) [33,34], analytic hierarchy process (AHP) [35], the Outranking approaches [36], goal programming, etc. An implementation of the decision made may be (and may not be) planned, and action ensued may have impact on the problem environment and will be re;ected on the perceptions of the participants on the problem. Feedback in fact can be gained at many moments of the decision making procedure, and there is always possible to reiterate a process that has been done. This situation is indicated in the @gure with two-way arrowhead lines. 3. Primary system functions and DSS classes For the domain analysis of DSSs for MCDM in natural resource management, comprehensive system requirement speci@cations are needed. DSSs for MCDM in natural resource management usually need to support group decision making and have several system users with multiple levels. Based on the identi@cation of the actors for the macro system for the decision problem solution [37] as well as past experiences in natural resource management [38,39], users for GDSSs for natural resource management are usually identi@ed as including a system administrator, a facilitator/analyst, stakeholders, domain experts, observers, and decision implementation agents. A system administrator is a real world person administrating the system. A facilitator/analyst is a real world person or persons facilitating the decision making processes in the system. Stakeholders are here de@ned to be any real world people whom decision makers, domain experts and the facilitator/analyst recognise as essential participants in the decision making process. Domain experts are the real world people who have professional expertise in the problem domain, e.g., economists, hydrologists, sociologists and environmental scientists. Observers are the real world people who want to study the system or the problem concerned for various reasons. Decision implementation agents are real world people or organisations who are going to implement the decision or recommendation resulted from the use of a DSS. The characteristics of these users and their interactions in decision making by using the DSS are distinctive aspects in GDSSs for MCDM in natural resource management, as shown in Fig. 2, which indicates the context of such a GDSS. In the context diagram for GDSSs for MCDM in natural
990
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999 Decision Problem Context Problem Definitions
System Administrator
Gained Insights
Observers
System Configuration
Feedback Comments
System Status Feedback
Feedback
Facilitator/ analyst
Domain Experts
DSS Identification Facilitation Imported Data
Exported Data
Feedback Stakeholders
Identification Judgements
Knowledge Identification Decision or Recommendation Comments
External Systems
Decision Implementation Agents
Fig. 2. Context diagram of the DSS.
resource management shown in the @gure, the DSS under consideration is the core of the context diagram with the decision problem context in the background. Data and information input to and output from the system are described brie;y along the directed arrows between actor rectangles and the shadowed system block. All system actors, which are roles played by users or other physical objects, are represented as classes in the form of rectangles. Examining the context diagram and modelling the decision making procedure for MCDM problems of natural resource management, as discussed in Section 2, can lead to the development of primary system functions that capture the main DSS requirements for MCDM in natural resource management. System functions of GDSSs for MCDM in natural resource management are represented by a collection of use cases [40–43], each of which is a short title for a system function. It is very important to ensure that the evaluation principles of system performance are met by GDSSs for MCDM in natural resource management since the success of a system can usually be assessed by using these principles. Evaluation principles in the literature [44–50] have been examined in order to identify the use cases relevant to the requirements of system performance. The collection of use cases that capture the primary requirements for general GDSSs for MCDM in natural resource management are listed in the tables of the appendix (Tables 1–4), each representing a system actor and its use case set. The use cases in the appendix mostly deal with group coordination and group decision making with con;icting objectives, which are essentially the goal of MCDM in natural resource management. After the examination of the use cases included in the appendix, primary classes of decision elements and other system components for DSSs for MCDM in natural resource management are identi@ed together with their de@nitions [17,19,40]. These classes are called primary because more system and platform speci@c classes, which may be di6cult to be generalised into generic elements for the DSS model, will be found at the later stages of system analysis and design. These primary classes are divided into two categories. The @rst category includes the classes of decision elements,
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
991
while the second contains the classes of other general fundamental system building components. Decision elements are abstracted representations of the key choices which need to be made by decision makers, and are collected into classes containing similar features. Classes of DSS building components are the material used by system developers to construct a DSS. Classes are regarded as the most fundamental units in object-oriented systems [17,19,40]. Primary system components of DSSs include the classes of decision elements de@ned above and other basic system building classes. Additional system component classes can be found when a speci@c DSS is developed with a programming language in a certain platform environment. 4. The DSS model and DSS development As discussed above, there is a need for the modelling of DSSs for MCDM in natural resource management. These DSSs deal with ill-structured strategic problems, leading to complexity in the development of the systems, and to heavy human–machine and human–human interaction in the systems. A DSS model mainly includes the outcomes of DSS domain analysis, i.e., the general DSS requirements, various classes, design of subsystems and system architecture, etc. The primary classes are organised into subsystems, which are then deployed in a system architecture before planning system implementation. Decision making in natural resource management involves various participants including multiple stakeholders, domain experts, and a facilitator. It is a problem of group decision making requiring support from a GDSS. Kreamer and King [51] de@ne GDSSs as interactive computer-based systems that facilitate the solution of ill-structured problems by a group of decision makers working together as a team. In the case of natural resource management, the “decision makers” are actually the governmental organisations or their agents, who may make the @nal decision about the resource allocation. The evaluation and analysis is often, however, undertaken by support sta0 or other stakeholders, who might need to make use of a GDSS to solve the problems in a group setting. The architecture diagram for the DSS model, which is essentially a GDSS, is shown in Fig. 3 together with the main subsystems. The subsystem of system administration administers the users and monitors the system. The problem understanding subsystem helps users understand the problem context and the decision making processes. The problem structuring subsystem deals with identi@cation of basic decision elements, such as attributes, criteria and uncertainties, during the brainstorming phase, the generation of alternatives, and the construction of criterion hierarchies. The evaluation subsystem elicits stakeholders’ judgements and @nally aggregates evaluations to obtain a result. The database subsystem manages documents to supply data and information for other subsystems. The user interface subsystem handles user commands and communications between the machine and the users. Fig. 3 shows the physical computing units in the system, their devices, and the communication links between the computing units. Computing units are capable of executing programs while devices can only support the execution of the programs. Subsystems of system administration, problem understanding, problem structuring and evaluation reside in the servers. They are accessed through the network access tools and operated via the user interface subsystem. To be precise, their codes are kept in the servers while they are operational on the front-end machines. They can be downloaded or
992
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999 Person Computer
Person Computer
Work Station
User Interface
Work Station
...
System Administration
Problem Understanding
Problem Structuring
Evaluation
Network
Server
...
Server
DataBase
DataBase
DataBase
DBMS
DBMS
DBMS
Fig. 3. System architecture.
accessed via other means by client machines while databases are stored and can also be administrated in the servers. Stakeholders of a natural resource management problem may be geographically dispersed across a large area and even internationally. They need to be able to use a DSS whenever and wherever they want to. This can be achieved by the accessibility of network nodes in o6ces and homes and via wireless communication means. Stakeholders and other system users can access the system through the network without any constraints of geographical locations. The access to the network ensures the availability of a DSS. This is one of the basic di0erences between traditional and Internet DSSs. The DSS model built in the study provides a general framework for the analysis and design of DSSs for natural resource management. It also allows such a system to meet the particular requirements of the evaluation principles for DSSs for MCDM. All entities of decision problems, decision elements, and system building components are represented in the DSS model as classes. The interactions among various classes accomplish the functionality of decision making [16]. The DSS model helps determine the system architecture for the design of a speci@c system. 5. The development of a DSS for water resources planning In order to validate the DSS model proposed for MCDM in natural resource management, a DSS for water resources management in South Africa was developed under contract to the South
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
993
Fig. 4. The home page of the WRC DSS.
African Water Research Commission (WRC) by using the DSS model described here. The objective of this WRC DSS was to support the process of applying MCDM concepts to decision making for water resource management in South Africa. The system facilitates managerial decision making by providing tools, procedures, and data that add structure to the decision making processes. The MCDM method used in the WRC DSS is a multi-attribute value theory (MAVT) based approach called scenario-based policy planning (SBPP) [8,39,52]. The WRC DSS is a group DSS (GDSS) based on the Internet. Decision making can be carried out without geographical restriction (users need only Internet access). The system allows a group of decision makers working together as a team to share information interactively, to generate ideas and actions, to choose alternatives and to negotiate solutions. Interactive information appears in di0erent forms, including the most commonly used web page formats, such as HTML pages, to guide the entire procedure of system operations. Users need only follow the ;ow of information from the Internet in order to fully make use of the system. Fig. 4 shows the home page of the system under Internet Explorer (IE) 5.0. The WRC DSS supports most of the decision making processes: problem structuring, evaluation and aggregation. At the phase of problem structuring, the problem under consideration is identi@ed and de@ned in terms of criteria, alternatives, and other related data. These may be thought of as the modules of problem structuring. The evaluation and aggregation phases include the module to elicit subjective judgements or value functions for evaluating alternatives, and the module to elicit weights for measuring the trade-o0s amongst criteria. They also calculate the weighted value of each alternative. Finally, the sensitivity of the weighted value to weights is examined.
994
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
The main functions of the system deal with group decision making carried out by several decision stakeholders, which represent di0erent interests, and an analyst who acts as a facilitator, for a particular decision problem case. The major decision making tasks supported in the system for a stakeholder include communications, criterion identi@cation, alternative set examination and creation, criterion hierarchy tree construction, decision alternative evaluation, and criterion comparison and analysis aggregation. The major decision making support provided to an analysts/facilitator, who is a special system user that facilitates the decision making activities at the system level, include the creation of the background and foreground policy scenario sets [8], system level criterion weighting, system overall analysis aggregation, and examination of the evaluation and weighting of stakeholders. For more information, see [37]. The completed system is composed of a system orientation component and six subsystems [37]. Some of them, such as the system orientation, only contain web pages while most of them consist of some clusters of classes. System orientation is to help users get familiar with the system and help them with the operations of the system. The six subsystems are system administration, problem orientation (understanding), problem structuring, evaluation, user interface, and database. System administration is responsible for the administration of users such as stakeholders, facilitators and system administrators. The subsystem of problem orientation helps understand the problem case. The subsystem of problem structuring deals with problem structuring activities for users to identify (edit) criteria, send and receive messages, construct criterion value tree, and view scenarios (facilitator can build the scenario set while normal decision participants may add extra scenarios). The subsystem of evaluation allows users to compare scenarios according to the criteria, and to weight criteria and aggregation. The subsystem of user interface contains various interface facilities such as windows, menus, dialogs, etc. The subsystem of database enables data to be stored and manipulated. The system analysis and primary design of the WRC DSS were carried out based on the DSS model developed in this study. System requirements and relevant classes were obtained from the existing documentation of the DSS model. Speci@cations of system requirements were captured with use cases, which need to be ful@lled by classes and class interactions. Various classes were thereafter de@ned based on the DSS model. The design of the system was then started. The subsystems and the system architecture of the WRC DSS was designed based on the DSS model by examining the classes identi@ed and after checking some speci@c technical implementation considerations. After the system analysis and design, the actual work of coding of the system was carried out in Java. The system was developed with an e0ort of about four person-months. The development e0ort is relatively little and therefore e6cient for a system with a scale of about 20,000 source lines of code [53,54]. Three decision analysis workshops [55] were conducted, by using a computerised DSS based on the WRC DSS but with some slight modi@cations in order to cater for the new situation, on a hypothetical decision problem (but one which was the subject of much debate locally), which was to consider strategies for the removal of alien vegetation from Table Mountain in Cape Town, South Africa. The workshop settings, including the material provided, the DSS used, and the decision making procedure followed, were the same for all three workshops. In the three workshops, di0erent groups of students were chosen in order to examine how the decision making methodology and the system could facilitate the decision analysis procedures for di0erent groups of people.
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
995
The decision problem is a di6cult and complicated natural resource management problem. In addition the workshops were expected to produce some insightful solutions in a limited time period. However, the outcomes of all three workshops demonstrated a considerable level of consistency and insights. Questionnaires were completed by subjects at the end of the workshops to evaluate the performances of the object-oriented MCDM methodology and of the DSS used in the workshops. It is clear that the respondents had a high level of satisfaction with regard to the various aspects of decision making stated in the questions of the questionnaires. It is concluded [55] that the DSS is e0ective in providing computerised decision making support, and that the methodology and the system can produce solutions which the participants have con@dence in. 6. Conclusions DSSs for MCDM in natural resource management are modelled in this paper with an object-oriented approach, which provides a uniform framework for both MCDM and DSS development. A general statement of system requirements for DSSs has been conceptualised to provide a set of core requirements and behaviour for DSSs for MCDM in natural resource management. Classes of decision elements for the analysis of decision problems and of other DSS components are identi@ed. In fact, the basic system building material includes these two kinds of classes, which are modelled with classes of object orientation for the purpose of decision facilitation and DSS development. After the identi@cation of classes, subsystems of DSSs are then designed, and a general system architecture is proposed. The DSS model designed in this paper may o0er an e6cient and e0ective way for the development of DSSs for MCDM in natural resource management by reusing generic system analysis and design results. The primary system requirements and classes of a speci@c system can be obtained from the existing DSS model. The system subsystems and system architecture can also be derived from the existing design results of the DSS model. A practical system was developed as part of a project for the South African Water Research Commission (WRC), based essentially on the DSS model as described above. It has demonstrated both e6ciency of the development process and e0ectiveness of the system developed. The system can be made available to public access through the Department of Statistical Sciences at the University of Cape Town. Future research may be focused on reuse in the form of existing commercial, proprietary, or home-developed documentation as well as class libraries, which are collections of classes stored in an organised way. The DSS model built in the paper for DSSs for MCDM in natural resource management is the @rst and essential step towards the development of reusable documentation and DSS class libraries. Generic aspects of DSS analysis and design are documented and de@ned in object orientation, and can then be implemented with speci@c programming languages. These classes are thereafter organised to form class libraries, which are then used to speed up the construction of speci@c DSSs, resulting in e6ciency in system development.
996
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
Appendix A. Summary of DSS system requirements Table 1 Summary of system requirements (facilitator/analyst) Facilitator/analyst & use cases
Brief description
Con@guration of user privileges Send messages Retrieve messages Guide for brainstorming of the problem
User privilege con@guration Message broadcast and mail Message reception and reading Guide for decision problem perceptions by stakeholders and domain experts Guide for generation of alternatives by stakeholders and domain experts Guide for construction of criterion hierarchies by stakeholders and domain experts Guide for uncertainty expression and management during various processes Guide for criterion relationship expression and exploration Check and control of the progress of the decision making Initial data, information and structure Selection of an overall alternative set for evaluation by stakeholders Construction of an overall criterion hierarchy Comparison of the overall criteria and @nal aggregation of all scores Sensitivity analysis of aggregation results Input small and large decision models Automatic information of decision making activities Textual and graphical problem document display Previous case demonstration Textual and graphical report of the decision analysis result
Guide for generation of alternatives Guide for construction of criterion hierarchies Guide for uncertainty expression and management Guide for expression of criterion relationships Co-ordination of decision making activities Initialisation of relevant data Generation of alternatives for evaluation Construction of the system level criterion hierarchy Evaluation of overall criteria and @nal aggregation Sensitivity Analysis Decision model input Decision making guidance Problem orientation Previous case examination Analysis Result Report
Table 2 Summary of system requirements (stakeholder) Stakeholder & use cases
Brief description
Send messages Retrieve messages Brainstorming of the problem Construction of a criterion hierarchy Generation/modi@cation of alternatives Evaluation of alternatives Evaluation of criteria and individual aggregation Sensitivity analysis Uncertainty expression and management Expression of criterion relationships Decision making guidance Problem orientation Previous case examination Analysis result report
Message broadcast and mail Message reception and reading Brainstorming problem perceptions Construction of an individual criterion hierarchy Generation/modi@cation of individual alternatives Evaluation of the selected alternatives by the facilitator/analyst Comparison of the criteria and individual aggregation of scores Sensitivity analysis of aggregation results Expression and exploration of uncertainty during various processes Expression and exploration of criterion relationships Automatic information of decision making activities Textual and graphical problem document display Previous case demonstration Textual and graphical report of the decision analysis result
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
997
Table 3 Summary of system requirements (domain expert) Domain expert & use cases
Brief description
Send messages Retrieve messages Brainstorming of the problem Construction of a criterion hierarchy Generation/modi@cation of alternatives Evaluation of alternatives Evaluation of criteria and individual aggregation Sensitivity analysis Check of decision elements brainstormed Check of alternatives generated Check of the overall and individual interest criterion hierarchies Uncertainty expression and management Expression of criterion relationships Value threshold setting Decision model input Decision making guidance Problem orientation Previous case examination Analysis result report
Message broadcast and mail Message reception and reading Brainstorming of problem perceptions Construction of an individual criterion hierarchy Generation/modi@cation of an individual alternative set Evaluation of de@ned alternatives Comparison of the criteria and individual aggregation of scores Sensitivity analysis of aggregation results Examination of decision elements brainstormed Examination of generated alternatives Examination of criterion hierarchies Expression and exploration of uncertainty during various processes Expression and exploration of criterion relationship Setting of value thresholds for action elements Input of small and large decision models Automatic information of decision making activities Textual and graphical problem document display Previous case demonstration Textual and graphical report of the decision analysis result
Table 4 Summary of system requirements (other system actors) Actor and use cases
Brief description
System administrator Registration of users Monitoring of system status
User registration System status monitoring
Implementation agents Send messages Retrieve messages Decision implementation Analysis result report
Message broadcast and mail Message reception and reading Implementation planning of the decision made Textual and graphical report of the decision analysis result
Observers Send messages Retrieve messages Trial use of the system as a stakeholder, a facilitator/analyst, or a domain expert External systems Exportation of data to the DSS Importation of data from the DSS
Message broadcast and mail Message reception and reading Simulation of a stakeholder, a facilitator/analyst, or a domain expert
Data exportation Data importation
998
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
References [1] Brill ED, Chang SY, Hopkins LD. Modelling to generate alternatives. Management Science 1982;28:221–35. [2] Bana e Costa CA, editor. Reading in multiple criteria decision aid. Berlin: Springer Verlag, 1990. [3] Chankong V, Haimes YY, Thadathil J, Zionts S. Multiple criteria optimisation: a state of the art review. In: Chankong V, Haimes YY, editors. Decision making with multiple objectives. Berlin: Springer-Verlag, 1985. p. 36–90. [4] Steuer RE. Multiple criteria optimisation: theory, computation and application. New York: Wiley, 1986. [5] Stewart TJ. A critical survey on the status of multiple criteria decision making theory and practice. OMEGA The International Journal of Management Science 1992;20(5/6):569–86. [6] Zionts S, Lot@ V. Recent development in multiple criteria decision making. OriON 1989;5:1–23. [7] Belton V, Stewart TJ. Multiple criteria decision analysis: an integrated approach. Boston: Kluwer Academic Publishers, 2002. [8] Stewart TJ, Scott L, Iloni K. Scenario based multicriteria policy planning for water management in South Africa. Report to the Water Research Commission by the Department of Statistical Sciences. WRC Report No 296/1/93, University of Cape Town, Cape Town, 1993. [9] Ellis H. Acid rain: operations research joins the battle to control a serious ecological threat. OR/MS Today 1992; August, 26 –28. [10] Hallefjord A, Jornsten K. A long range forestry planning problem with multiple objectives. European Journal of Operational Research 1986;26:123–33. [11] Bhargava HK, Sridhar S, Herrick C. Beyond spreadsheets: tools for building decision support systems. Computer 1999;32(3):31–9. [12] Batachia IL. Towards user-centred and cost-e0ective development of environmental decision support systems. Studies in Informatics and Control 1999;8(4):279–93. [13] Rao VS, Jarvenpaa SL. Computer support of groups: theory-based models for group decision support systems (GDSS). Management Science 1991;37(10):1347–62. [14] Wang S. Decision support systems domain analysis and modelling: an object-oriented technique. Information and Systems Engineering 1995;1:289–301. [15] Schmoldt DL, Perterson DL. Analytical group decision-making in natural-resources-methodology and application. Forest Science 2000;46(1):62–75. [16] Liu D, Stewart TJ. Integrated object-oriented framework for MCDM and DSS modelling. Submitted for publication, 2002. [17] Booch G. Object-oriented design with applications. California: Benjamin Cummings, Redwood City, 1991. [18] Gauthier L, Neel T. SAGE: an object-oriented framework for the construction of farm decision support systems. Computers and Electronics in Agriculture 1996;16(1):1–20. [19] Graham I. Object-oriented methods, 2nd ed. New York: Addison-Wesley, 1994. [20] Missiko0 M. An object-oriented approach to an information and decision support system for railway tra6c control. Engineering Applications of Arti@cial Intelligence 1998;11(1):25–40. [21] Muhanna WA. An object-oriented framework for model management and DSS development. Decision Support Systems 1993;9:217–29. [22] Bomme P, Zimmermann T. An intelligent symbolic object for decision making procedures. In: Topping BHV, editor. Developments in computer aided design and modelling for civil engineering. Edinburgh: Civil-Comp Press, 1995. p. 1–11. [23] Chen H, Sinha D. An inventory decision support system using the object-oriented approach. Computers & Operations Research 1996;23(2):153–70. [24] Dolk DR, Konsynski BR. Knowledge representation for model management systems. IEEE Transactions on Software Engineering 1984;10(6):619–28. [25] Fedra K, Jamieson DG. An object-oriented approach to model integration: a river basin information system example. In: Kovar K, Nachtnebel HP, editors. Application of geographic information systems in hydrology and water resource management. Wallingford: International Association of Hydrologic Science, 1996. p. 669–76. [26] Hagmann C. An object oriented design for local area decision network (LADN) model for small group decision support systems (SGDSS). PhD dissertation, Kansas State University, 1989.
D. Liu, T.J. Stewart / Computers & Operations Research 31 (2004) 985 – 999
999
[27] Lal H. Object-oriented decision support system in logic programming (for farm machinery). In: Zobrist GW, Leonard JV, editors. Progress in simulation, vol. 1. New Jersey: Norwood, 1992. [28] Le Claire C. Decision support systems: an object-oriented conceptual architecture. PhD dissertation, Oklahoma State University, 1989. [29] Lenard ML. An object-oriented approach to model management. Decision Support Systems 1993;9:67–73. [30] Pillutla SN, Nag BN. Object-oriented model construction in production scheduling decisions. Decision Support Systems 1996;18:357–75. [31] Rizzoli AE, Davis JR, Abel DJ. Model and data integration and re-use in environmental decision support systems. Decision Support Systems 1998;24(2):127–44. [32] Phillips LD. People centred group decision support. In: Doukidis G, Land F, Miller G, editors. Knowledge based management support systems. Chichester, UK: Ellis Horwood, 1988. p. 208–24. [33] Keeney RL, Rai0a H. Decisions with multiple objectives: preferences and value tradeo0s. New York: Wiley, 1976. [34] Von Winterfeldt D, Edwards W. Decision analysis and behavioural research. Cambridge: Cambridge University Press, 1986. [35] Saaty TL. The analytical hierarchy process. New York: McGraw-Hill, 1980. [36] Roy B. The outranking approach and the foundations of ELECTRE methods. In: Bana e Costa CA, editor. Readings in multiple criteria decision aid. Berlin: Springer, 1990. p. 155–83. [37] Liu D. An object-oriented approach to structuring multicriteria decision support in natural resource management problems. PhD Thesis, University of Cape Town, 2001. [38] Stewart TJ, Joubert A, Liu D. Group decision support methods to facilitate participative water resource management, report to the water research commission by the Department of Statistical Sciences. WRC Report No 863/1/01, University of Cape Town, Cape Town, 2001. [39] Stewart TJ, Joubert A, Scott L, Low T. Decision support procedures for water resources management, report to the water research commission by the Department of Statistical Sciences. University of Cape Town, Cape Town, 1996. [40] Booch G, Rumbaugh J, Jacobson I. The uni@ed modelling language user guide. Massachusetts: Addison-Wesley, 1999. [41] Gossain S. Object modeling and design strategies. Cambridge: Cambridge University Press, 1998. [42] Jacobson I. Object-oriented software engineering. Massachusetts: Addison-Wesley, 1992. [43] Martin L. Succeeding with the booch and OMT methods: a practical approach. California: Addison-Wesley, 1996. [44] Adams DA, Courtney Jr. JF, Kasper GM. A process-oriented method for the evaluation of decision support system generators. Information and Management 1990;19:213–25. [45] Applegate LM, Chen TT, Konsynski BR, Nunamaker JF. Knowledge management in organisational planning. Journal of Management Information Systems 1987;3:20–8. [46] Buede D. Superior design features of decision analytic software. Computers & Operations Research 1992;19(1): 43–57. [47] Eom SB, Lee SM, Kim EB, Somarajan C. A survey of decision support system applications (1988–1994). Journal of the Operational Research Society 1998;49(2):109–20. [48] Evans GE, Riha JR. Assessing DSS e0ectiveness using evaluation research methods. Information and Management 1989;16:197–206. [49] Le Blanc LA, Jelassi MT. DSS software selection: a multiple criteria decision methodology. Information and Management 1989;17:49–65. [50] Rizzoli AE, Young WJ. Delivering environmental decision support systems: software tools and techniques. Environmental Modelling & Software 1997;12(2–3):237–49. [51] Kreamer KL, King JL. Computer-based systems for co-operative work and group decision making. ACM Computing Surveys 1988;20(2):115–46. [52] Stewart TJ, Scott L. A scenario-based framework for multicriteria decision analysis in water resources planning. Water Resources Research 1995;31:2835–43. [53] Albrecht AJ, Ga0ney JE. Software function, source lines of code, and development e0ort prediction: a software science validation. IEEE Transactions on Software Engineering 1983;9(6):639–48. [54] Bassman M, McGarry F, Pajerski R. Software measurement guidebook, SEL-94-102. Software Engineering Laboratory, NASA/GSFC, 1995. [55] Liu D, Stewart TJ. Object-oriented MCDM for natural resource management. Submitted for publication, 2002.
