효과적인 단기 · 중기 · 장기 전략계획 수입을 위한 양방향 추론 인공신경망 모형에 관한 연구

다국어 초록 (Multilingual Abstract)

For the sake of effective strategic planning simulation, this paper proposes a bi-directional inferencing neural network model, named SPBINN (Strategic Planning Bi-directional Inferencing Neural Network). The model is composed of two major neural network sub-models: (1) forward inferencing model and (2) backward inferencing model. The forward inferencing model supports building short-term, mid-term, and long-term strategic planning in response to pre-determined inputs. The backward inferencing model, however, enables so called what-if analysis with which decision makers are able to investigate which kind of conditions make a specific strategic goal be acquired satisfactorily. We adopt a popular backpropagation learning algorithm to train both neural network sub-models, but the solution process of SPBINN is unique in the sense that conflict resolution mechanism helps resolve conflicts between short-term, mid-term, and long-term strategic planning. Experiments with real data of Korea Cosmetic Industry showed that SPBINN, a proposed neural network-based strategic simulation model, can provide robust strategic planning results under conflicting situations and help decision-makers to choose the most promising strategy.