基于抽样方法的不确定极限学习机

实际应用中的大量数据具有不确定属性,而传统的挖掘算法无法直接应用在不确定数据集上.针对不确定数据的分类问题,提出一种基于抽样方法的不确定极限学习机.该算法通过抽样的方法,对不确定数据集中样本的抽样实例进行学习和分类,得到该不确定样本的所属类别的概率,从而实现了传统极限学习机分类算法对不确定数据的分类,并极大降低了不确定对象实例的枚举代价.实验结果表明,该算法在不确定数据的分类问题中具有较好的有效性和高效性.

Sampling Based Uncertain Extreme Learning Machine

Large amounts of data in real-world applications have inherent uncertainty. Traditional learning algorithms cannot be applied directly onto uncertain datasets. Aiming at classification problems over uncertain data， a sampling based uncertain ELM(extreme learning machine) was proposed. Instances were first sampled out of uncertain objects， and then learnt with uncertain ELM. The uncertain objects would be assigned to their classes respectively according to the probabilities aggregation method. The classification was realized by the proposed algorithm in this paper over uncertain data avoiding the enumeration of instances. The experimental results indicated the efficiency and effectiveness of our algorithm.