Agent Oriented Requirements Engineering for a Data Warehouse Manoj Kumar 1 , Anjana Gosain 2 , Yogesh Singh 3 1 Ambedkar Institute of Technology, Delhi, India Department of Computer Science & Engineering 2, 3 Guru Gobind Singh Indraprastha University, Delhi, India University School of Information technology manojgaur@yahoo.com, anjana_gosain@hotmail.com, ys66@rediffmail.com Abstract In recent years, a number of requirements engineering (RE) proposals for a data warehouse (DW) systems have been made. In the traditional/operational systems, requirements engineering has been divided into two phases: early & late requirements engineering phase. Most of the data warehouse requirements engineering (DWRE) approaches have not distinguished early requirements engineering phase from late requirements engineering phase. A very few approaches are seen in the literature that explicitly model early & late requirements for a DW. In this paper, we propose an AGDI (Agent-Goal-Decision-Information) model to support early requirements engineering issues for a data warehouse. Here, early requirements have been modeled through organization modeling and goal modeling activities as an illustration of proposed AGDI model to support decisional goals of the organization for which DW is to be built. Keywords: agent, early requirements engineering, late requirements engineering, data warehouse 1. Introduction In the last decade, great interest has been shown in the development of Data Warehouses (DW). The initial thrust of DW was in decision making that principally involved numeric facts and textual dimensions. In recent years, however, Data Warehouses have been proposed for domains of image data [14], voice data [10]. For DW in the decision-making domain, two different approaches, data-driven [8] and requirements-driven [2] have been used. In the former, data is gathered from operational systems into DWs whereas in the latter the attempt is to identify the information needs to be met by the DW. In these approaches, the real issue is that of DW design: given data needs what is the logical structure of the DW. Proposals to add a conceptual layer on top of the basic data layer have been made, for example, by Jarke et al. [9]. Here, it is assumed that the conceptual objects can be determined but the question of what are useful conceptual objects for a DW and how these are to be determined is not addressed. To answer this, we need an explicit Requirements Engineering (RE) phase in DW development. The requirements engineering task has been divided into two phases: early requirements engineering phase and late requirements engineering phase [16, 17]. The “early-phase” of requirements engineering activities include those that consider how the intended system would meet organizational goals, why the system is needed, what alternatives might exist, what the implications of the alternatives are for various stakeholders, and how the stakeholders’ interests and concerns might be addressed. The emphasis here is on understanding the “whys” that underlies system requirements [15], rather than on the precise and detailed specification of “what” the system should do. The notion of agent and related mentalisic notions are used in all software development phases from early requirements analysis down to the actual implementation [4, 17, 18]. The various data warehouse requirements engineering (DWRE) approaches have not distinguished early requirements engineering phase from late requirements engineering phase. However, in [5] the early phase of Tropos [Bre04] has been extended to the requirements engineering of data warehouses. This approach [5] uses stakeholder dependencies, which are represented in an actor diagram. Thereafter, two perspectives are proposed, organizational and decisional. In the former, facts are identified and associated to goals of different actors. In the later, each fact is related to their dimensions and a set of measures is found out and associated with facts. However, this approach does not look at the decisional goals [12] of the organization for which DW is to be built. In [12], GDI (Goal-Decision-Information) model [11], has been viewed in two ways, one from the organizational and the other from the technical perspective. The former look upon the warehouse as embedded in an organization and considers the manner in which it supports organizational tasks. The latter deals with issues of data warehouse contents, their broad properties etc. This approach [12] is modeling the late phase requirements but not the early phase requirements in the context of DWs. In this paper, we propose an AGDI model as an extension to the GDI model [11] to model early requirements for a DW. Here, we introduce the notion of agent, which represent stakeholders of the organization for which DW is to be built. The various stakeholders’ dependencies have been modeled as goal, decision and information dependencies among agents. The proposed AGDI model supports early requirements modeling activities as organization modeling & goal modeling. The organization and goal modeling activities capture early requirements for a DW. Organization of the paper is as follows: The section 2 discusses the proposed AGDI model, which extends GDI model to support early requirements modeling. In section 3, proposed AGDI model is illustrated to model early requirements through an example of a university, for which DW is to be built, followed by conclusion in section 4. 2. AGDI (Agent -Goal- Decsion-information) Model for DW Requirements Modeling The GDI model [11] starts with the determination of goal of the organization with the help of decision makers by assuming that only decision maker are the stakeholders responsible for decision making activities in the organization. The other stakeholders and their dependencies for achieving the goals of the organization are also important to be modeled. For this GDI model does not have ACM SIGSOFT Software Engineering Notes Page 1 September 2009 Volume 34 Number 5 http://doi.acm.org/10.1145/1598732.1598737 DOI: 10.1145/1598732.1598737