Based on the analysis of the classification accuracy of the learning with local and global consistency(LLGC)
algorithm being influenced greatly by a suitable setting of parameters, a kind of barebones LLGC(BB-LLGC) algorithm with
less parameters is proposed. The objective function defined on a graph is simplified to make it not be influenced by parameter
alpha. During the label propagation process, only the predicted labels of unlabeled samples are propagated to its neighbors
according to a similarity metric, while the labels of labeled samples are kept unchanged so as to ensure the correctness of
the source of label propagation. Experimental results concerning on several UCI datasets show that, compared with LLGC,
the BB-LLGC has advantages of less control parameters, simple operation procedure, high classification accuracy and fast
convergence speed.