An adequate decision making is heavily dependent on data fusion processes. The only way a decision making agent can infer a decision which is adequate to the current state of an environment is through gaining a situation awareness regarding relevant aspects of it which is achieved through data, information, and knowledge fusion. The following paper covers some of the latest proposals for frameworks and architectural approaches to building fusion-based decision making systems, both formal and conceptual. The article also proposes a new architectural approach to building more extensible fuison-based systems through adding an explicit prediction block. The motivation for that architectural solution was the blistering pace of learning-based prediction systems development witnessed by scientific and engineering communities during the last decade which brought us a plethora of well-established methods and methodologies.