Local decision making and decision fusion in hierarchical levels

Abstract

Decision-making, Decision-fusion, Neural networks, Classification, Machine learning, 68T01, 68T05, 68T30, 68T37, 68T40,

A new systematic and flexible method for developing hierarchical decision-making models

“…Details of the case study of 'Determining the fault degree in a specific tank in a 4-tank water circulation system' and the applied alternative DM model (local DM in multiple levels) to solve the same case study can be found in [14,19]. This system is made up of 4 water tanks and a reservoir, as shown in Figure 5 [14,19]. The reservoir feeds the tanks via 2 water pumps.…”

“…The states of the system due to different fault configurations and the actual states of the system when the system is error-free are compared with each other. After some postprocessing, the input data (called normalized error data in [14]) are obtained. Each input data element has 2 components.…”

“…In [18], a complex structured DM model was put forward, which is a hierarchically organized DM framework to represent complex structured knowledge. Developing agent structures that function as experts for DM problems is also an important area of investigation for researchers [14,19]. Machine learning tools and heuristic search methods like genetic algorithms (GAs) [20] are widely employed to develop agent structures.…”

“…However, as the complexity of the application increases, it is difficult and inefficient to provide the organization of experts so effortlessly. Dealing with a huge amount of data coming from different observations [11]; the necessity of task planning, gathering, and evaluating information from different heterogeneous sources [12]; the essentiality for applicability of a learning paradigm in different frameworks in order to handle DM problems more accurately [13]; and the comfort in establishing organization between agents without difficulty [14] are the typical reasons to use hierarchical and multilevel frameworks. In multilevel DM frameworks, experts in any level have a direct or indirect impact over the actions and decision of the experts' at other levels.…”

“…In this study, a practical solution for these research questions is sought. The main aim of this study is to establish a new multilevel DM system architecture as in [14] for handling DM problems with high complexity. In this architecture, the main aspiration is to obtain and utilize agents with increasing complexity in a hierarchical manner for the DM task(s) and carry on this procedure until the performance of the agents is sufficient.…”