ABSTRACT
Neural network is a technique of articial intelligence widely used in pattern recognition and machine learning. It possesses strong adaptability and learning ability, and its precision in the data classication and prediction is extremely high. Being dierent from classical statistical methods, the neural network can be applied to data analysis of small sample, without satisfying normal distribution. Because of the nonlinear transformation involved in the radbas layer, neural networks can work well in imitating complex data. However, neural networks are very complex in the computation, and the time required by their learning algorithms is in most cases greater than that required by other methods in articial intelligence. Besides, it cannot provide a crisp description of its discovery pattern. Variable precision rough set (VPRS) has the advantage that we can get learning probability decision rules from imprecision data and that it can be expressed by the customer’s understanding. e hybrid approach of VPRSs, fuzzy sets, and neural network can shorten the time of network learning and improve the prediction ability of data and strengthen explanatory power.
