It is believed that a data warehouse is for operational decision making. Recently, a proposal was made to support decision making for formulating policy enforcement rules that enforce policies. These rules are expressed in the WHEN-IF-THEN form. Guidelines are proposed to elicit two types of actions, triggering actions that cause the policy violation and the corresponding correcting actions. The decision-making problem is that of selecting the most appropriate correcting action in the event of a policy violation. This selection requires information. The elicited information is unstructured and is “early.” This work is extended by proposing a method to directly convert early information into its multi-dimensional form. For this, an early information mode is proposed. The proposed conversion process is a fully automated one. Further, the tool support is extended to accommodate the conversion process. The authors also apply the method to a health domain.
TopIntroduction
Traditionally a data warehouse (DW) supports operational work related decision-making (Inmon, 2005). Recent proposals address the full range of corporate decision making. (Prakash and Prakash, 2015) address the issue of providing support for policy formulation decisions. (Prakash and Gupta, 2014) support decision making for formulating policy enforcement rules (PER).
(Prakash, 2010) showed that there is in fact a decision continuum that exists in the decision making environment of an organization. The outer most layer of the continuum is where policy formulation decisions are taken. Once policies have been formulated, policy enforcement rules are formulated. PER formulation decisions form the next inner layer. PERs enforce policies in the organization. Once the policy enforcement rules are formulated, operational decisions are taken. Operational decision form the inner most layer of the continuum. It is possible to move from the policy decision layer to the PER layer and from the PER layer to the operational layer. A DW is required to support the three layers of decision making.
Based on this view, the data warehouse requirements engineering (DWRE) process has three main steps (a) identifying the set of decisions, (b) eliciting information to support the decisions, and (c) converting information into multi-dimensional structure.
A number of techniques exist to convert information to multi-dimensional structure. They can be classified based on the number of steps required to complete this conversion. The one-step approach identifies facts and dimensions as a one shot activity. The two and the three step approaches break the requirements engineering task into smaller pieces and therefore techniques are developed for each piece.
Techniques of (Giorgini, 2005; Giorgini, 2009; Salinesi and Gam, 2006; Mazon, 2007) follow the one step approach where the moment information is determined, by the stakeholder, facts and dimensions are identified. This process of identification is unguided and completely based on the experience of the requirements engineer/developer.
Two-step approaches have been proposed by (Boehnlein, 1999; Boehnlein, 2000; Bonifati, 2001; Corr and Stagnitto, 2012). The first step in the two-steps is usually a generic representation of information. (Boehnlein, 1999; Boehnlein, 2000) map service measures, from stakeholders, to SERM. In the next step multi-dimensional structures are obtained from SERM diagram. In the case of (Bonifati, 2001), abstraction sheets capture quality, variation factors among others. This is treated as information which is converted to the MD structure in the second step. (Corr and Stagnitto, 2012) obtains MD structure from tables. It is in the tables that information is captured. In all the three techniques, the intermediate step can be used to guide the process of identifying the MD structure.