The paper proposes a hybrid approach of grey rough set and probabilistic neural network for uncertain decision. Grey rough set model is tolerant of noise. By setting a level of grey degree, redundant attributes are eliminated from decision table, a minimal knowledge representation is derived and the set of rules are generated through the grey rough set model. Subsequently, the reduced decision table is forwarded to probabilistic neural networks for classification and decision. The additional properties to PNN provided by the grey rough set analysis are input dimensionality reduction by the elimination of irrelevant features, a fast learning process, explanation facilities providing, hidden patterns finding in data and uncertainty treatment. The research result reveals that the hybrid approach has a high accuracy in classification and decision. The method can be applied to uncertain decision with ambiguous, incomplete and noisy database.
展开▼
