Abstract:To better impact the training model with the inherent distribution of the training dataset, a twin support vector regression called density-weighted twin support vector regression is proposed. Firstly, the density-weighted value is computed based on the ??-nearest neighbor algorithm. Then, the values of density-weighted are introduced to the standard twin support vector regression. It is found that the proposed algorithm can well reflect the inherent distribution of the
training dataset and lead to a more accurate impact on the training model. Experimental results on six UCI datasets show the effectiveness of the proposed algorithm.