The role of technological forecasting is not merely to predict "what will happen and when", but is to aid R&D planning and management in peering into the future. Although Delphi survey technique has been found quite useful for long-range technological forecasting, from the user's point of view (such as R&D planners and project managers), the conventional technique applied to predict occurrence dates of closely related future technology developments has a major shortcoming. Namely, the Delphi result can be used only in exploratory forecast by individual base. Since the technique does not consider the significance of relationship and interaction between developments, its result can not be interpreted easily in normative aspect of forecasting, i.e., impact analysis associated with user's planning context, or simulation work under various condition regarding R&D policy or plan. Moreover, existing supporting tools concerned with interaction, e.g. Cross-impact analysis, requires the burdensome explicit information on interrelationships.
This study, following the line of Cross-impact analysis, is concerned with the interrelationships among technology developments over time horizon, but seeks to avoid imposing any additional burden. Instead, it utilizes the information apparently skipped in the original Delphi survey data, by adopting a knowledge-based system approach.
From the user's point of view, the Delphi data can be considered extensive knowledge of experts on a specific domain, spending much time and cost. It will be useful to convert one-time forecast of the Delphi into a knowledge-base of decision support system(or expert system) and to refer to the experts' knowledge-base at instance when a user has his own information or situationspecific evidence.
To construct a knowledge-base with focus on interrelationship, knowledge representation and acquisition scheme is required. And it should be designed to describe the real forecasting domain, i.e. technological progress. The future technological progress has uncertain, complex and dynamic feature in its direction and rate of change. To capture complex and dynamic interaction this approach adopts a dynamic system model of semi-Markov process.
This approach is backed up practically by the characteristics of Delphi data. Incidentally, when panelists are asked to predict occurrence dates of multiple developments in a specific field, before they give the response for each development, it is natural for them to give implicit consideration to the interrelationship of developments in his perceptional map for future. As far as we skillfully elicit implicit information, a model considering interrelationship between developments can be based on original Delphi data without giving any additional burden for parameter assessment.
The major objectives of this research can be summarized as follow:
(1) Development of a decision support system (or knowledge-based system) for technological forecasting and impact analysis utilizing underlying information in Delphi data.
(2) Development of a knowledge representation and acquisition scheme considering uncertain, complex and dynamic technological progress.
First, the semi-Markov process approach is developed for sequential technological progress, by capturing its two probabilistic characteristics: the uncertain sequence of developments and the uncertain time length between successive developments. The stochastic natures of the technological process is represented by the "state transition" and the "state holding time" concepts in a semi-Markov process.
This study generalizes the semi-Markov model for a complex technological progress showing fusion and specification. The approach focuses on the cluster structure defined by interdependence measure. In addition, in order to improve utilization of underlying information, a new weighting technique based on the sample variance and covariance is developed for opinion pooling the non-homogeneous Delphi panel groups.
Finally, the approach is implemented to a prototype knowledge-based system, named DESMA(DElphi Supporting System using Semi-Markov Approach). The system provides a flexible and useful forecasting and post-analysis tool for R \& D planners or project managers to give insights on the dynamics of future technological progress.
This study suggests a new type of integrated forecasting approach, in the sense that, whose objective is utilization of exploratory forecast data in normative side of planning and management process. The development of a realistic knowledge representation model describing future technological progress and construction of practical decision support system would be the further research.
技術革新의 가속화에 따라 정책입안이나 계획수립에 있어 기술적 진보에 대한 전망, 즉 기술의 발전방향과 속도 예측의 중요성이 높아지고 있으나 예측하고자 하는 기술진보는 기본적 성격으로서의 불확실할(Uncertain)뿐만 아니라 여러가지 기술이 복잡한 상호영향하에서(Complex), 시간에 걸쳐 동태적(Dynamic)으로 진보해 가는 양상을 띄고 있어 어려운 과제가 되고있다. 정보가 적고 불확실성이 높은 중·장기 기술예측의 방법으로서는 전문가의 주관적 의견을 조직적으로 수렴하는 Delphi 기법이 널리 활용되고 있으나, 종래의 방법은 상호 영향을 무시한 개별기술의 출현 가능성 및 출현 시기예측에 대한 탐색적(Exploratory) 예측에 그치고 있어 투입되는 시간과 노력에 비해 효용도가 낮은 단점이 있다. 사용자(정책 입안자 또는 계획수립자)의 입장에서는 기술전망의 중요성은 일시점의 예측 자체보다는 정책 또는 계획과 관련한 당위적(Normative)예측 및 상황변화에 따른 영향분석(Impact analysis)이 보다 중요하기 때문이다.
따라서 본 硏究는 Delphi 자료를 '일회용 당위적 예측을 위한 전문가들의 주관적 판단 Data' 라기 보다는 '장기적으로 활용하여야 할 전문가들의 지식베이스(Knowledge-base)'라는 새로운 각도에서 조명하여 사용자의 입장에서 다양한 상황적 정보와 함께 수시로 기술전망 및 정책관련 영향분석을 할 수 있는 방법론과 시스템을 개발하고자 한다.
이를 위하여 연구의 주안점을 기존의 Delphi data가 내포하고 있는 묵시적인 연관관계 정보를 지식베이스화 할수있는 지식표현 및 추출체계에 두고 기술진보과정의 불확실성, 복잡성, 동태성을 동시에 고려한 동태적 시스템모형을 도입하였다. 먼저 관련성이 높은 일정분야의 기술진보에 대해서는 순차적인 연관관계에 촛점을 맞춘 Semi-Markov Process 모형을, 보다 복잡한 전개 양상을 보이는 진보과정에 대해서는 순차성과 함께 Cluster관계를 동시에 고려한 Semi-Markov network 모형을 통해 일련의 기술들의 동태적 상호연관관계를 표현 추출했다. 또한 여러 Delphi group들의 추정치 분포가 갖는 정확도와 정보중복을 고려한 가중치를 추출하여 Data Pooling을 할 수 있는 기법을 개발하였다. 나아가 이러한 방법론을 실용화하기 위해 추출된 정보를 활용해 예측 및 영향분석 시뮬레이션을 할 수 있는 지식베이스 시스템인 DESMA를 구축하였다.
