Increased evidence indicates that long non-coding RNA (lncRNA) plays a vital role in intricate human diseases. Nonetheless, the current pool of identified lncRNAs linked to diseases remains restricted. Hence, the scientific community emphasizes devising a reliable and cost-effective computational approach to predict the probable correlation between lncRNAs and diseases. It would facilitate exploring the underlying mechanisms of lncRNAs in ailments and generating novel disease treatments. In this study, we propose a novel approach for predicting lncRNA-disease associations through automatic meta-path graph generation (AMPGLDA). Firstly, we integrate the multi-view information of lncRNAs, diseases, and miRNAs to construct a heterogeneous graph. We then extract global features from nodes. Based on this heterogeneous graph, AMPGLDA generates multiple meta-path graph structures through a meta-path automatic generation module and uses the feature extraction module to learn the feature representations of lncRNAs and diseases from meta-path graphs. Ultimately, AMPGLDA is able to accurately predict the association between lncRNA and disease by utilizing a deep neural network classifier. The AMPGLDA model achieves impressive results with AUC and AUPR scores were 0.9985% and 0.9984% under the test set, respectively, outperforming several advanced LDA prediction models. Furthermore, three case studies demonstrate how accurately it is able to find new associations.