The cognition-driven decision process for business intelligence : a model and techniques

Publication Type:
Thesis
Issue Date:
2009
Full metadata record
Files in This Item:
Filename Description Size
Thumbnail01Front.pdf4.19 MB
Adobe PDF
Thumbnail02Whole.pdf87.76 MB
Adobe PDF
NO FULL TEXT AVAILABLE. This thesis contains 3rd party copyright material. ----- Business intelligence (BI) has received increasing attention from both academics and practitioners during the past two decades. However, cognitive decision support remains weak in today's BI systems, as do other types of decision support systems (DSS), although it has long been recognized as an important consideration. In DSS, cognitive decision support is concerned with supporting decision makers from a human cognition perspective. Human cognition refers to mental processes such as thinking, sensing, comprehension and inferring. Confronted with decision situations, decision makers develop situation awareness (SA) based their mental models and other mental constructs during cognition. SA and mental models are considered to be two extremely crucial prerequisites for decision making, particularly in ill-structured and dynamic decision situations with uncertainties, time pressures and high personal stakes. In today's business domain, decision making is becoming increasingly complex. To be successful, managers tend to rely on their SA about their business environment. This research focuses upon cognitive decision support in the BI environment. Design research is employed as the research methodology in this research, during which a theoretical model was suggested, related information systems (IS) techniques were proposed, and IS artefacts were developed and evaluated. We proposed a cognition-driven decision process (CDDP) model incorporating the manager (decision maker or user) with his/her SA and experience (mental models) as its central component. The goal of the CDDP model is to facilitate cognitive decision support to the manager in his/her decision process on the basis of a data warehouse (DW) to achieve better business intelligence. In the CDDP model, a typical decision process consists of a number of decision cycles. Each decision cycle includes five steps. (1) The manager describes and inputs his/her SA using natural language, from which SA semantics are extracted by the system; (2) The system retrieves relevant experience according to the SA; (3) The system generates the information need according the retrieved experience; (4) The system retrieves situation information from the DW according to the information need; (5) The system presents the retrieved situation information to the manager with the guidance of the experience; (6) the manager perceives the situation information and updates his/her SA. With the iteration of decision cycles, the manager's SA is accumulated and reinforced towards richer and higher level understanding about the decision situation, which eventually leads to a final decision. In the CDDP model, the computer does not generate specific decision recommendations or suggestions. Ironically, computers are intended for seeking situation information based on the manager's SA and past management experience. The retrieved situation information is closely related to the current decision situation and is used to support the manager to develop richer SA. The final decision to the current situation is more naturalistically made by the manager with high level SA. According to the CDDP model, we developed relevant IS techniques (methods and algorithms) to support all the six sub-processes of each decision cycle in a decision process. Using the IS techniques, we designed and developed a prototype system called FACETS according to the CDDP model. Experiments were conducted to evaluate the proposed techniques using FACETS as the test bed. FACETS per se was also evaluated as a complete system. The experiment results show that FACEST, as an implementation of the CDDP model, has significant implications for cognitively supporting business managers to deal with ill-structured problems.
Please use this identifier to cite or link to this item: