There are two significant problems with existing collaborative filtering (CF) models based on graph convolutional networks (GCNs). Most original graphs have noise and data sparsity problems that can seriously impair the model performance. In addition, for large user project graphs, the explicit message passing in traditional GCNs slows down the convergence speed during training and weakens the training efficiency of the model. A graph convolution collaborative filtering model combing graph enhancement and sampling strategies (EL-GCCF) was proposed to respond to the above two points. In the graph initialization learning module, the structural information and the feature information in the graph were integrated by generating two graph structures. The original graph was enhanced and the noise problem was effectively mitigated. Explicit message passing was skipped because of the multi-task constrained graph convolution. The over-smoothing problem in training was effectively mitigated and the training efficiency of the model was improved by using an auxiliary sampling strategy. Experimental results on two real datasets show that the EL-GCCF model outperforms many mainstream models and has higher training efficiency.