许多学习算法都存在这样一个偏置:属性集中的属性同等重要.然而,这种假设不一定实际.如果把属性集中的属性根据实际情况考虑为分别具有不同的重要性,那么由此获得的模型应该更合理,也有不少学者将此考入到算法中.文章将计算属性约简的问题转化为计算集合覆盖约简问题的思想,通过将描述用户需求或偏好的属性序纳入考虑,设计了基于用户需求的覆盖约简算法,并且对计算复杂性分析.最后运用实例验证了算法的可行性和有效性.%Many learning algorithm has such a bias:the attributes of the attribute set are equally important.However,this assumption is not reasonable,also not practical.If the attributes of the attribute set on according to the actual situation of the consideration for the importance of different respectively,so I suppose the model should be more practical.That transforms calculating the attribute reduction problem into the calculation covering reduction problem.Through this will describe the user needs or preference attribute sequence into consideration.The design based on user demand covering of reduction algorithm is proposed.And through the example,analysis the calculation complexity,expounds the feasibility and effectiveness of the algorithm.
展开▼