The study of the lncRNA-diseases association is a prevalent project. Researchers have developed many effective computational models to assist experimental research. In this paper, we proposed a computational model (GILDA) to predict the potential lncRNA-disease associations based on graphlet interaction. This model established lncRNA similarity graph and disease similarity graph. Based on the graphlet interaction, we calculated the numbers of graphlet interaction isomers in the similarity graphs respectively, which were then taken as features to train the model, and obtained the predicted lncRNA-disease association scores. Compared with the previous calculation models, GILDA fully considers the direct and indirect relationship between two nodes through the graphlet interaction. The AUC value of GILDA in global LOOCV is 0.8844, in local LOOCV is 0.8468, and in fivefold cross-validation is \(0.8742\pm 0.0057\). At the same time, we made case studies of hepatecellular carcinoma, renal cancer and adenocarcinoma. The results proved that among the first 10 lncRNAs predicted for the three diseases, the prediction accuracy was 100%, 100% and 80%, respectively.