Vector-borne diseases carried by mosquitoes, ticks, and other vectors are among the fastest-spreading and most extensive diseases worldwide, mainly active in tropical regions. Also, in the context of the current climate change, these diseases are becoming a hazard for other climatic zones. Hence, drug repurposing methods can identify already approved drugs to treat them efficiently, reducing development costs and time. Knowledge graph embedding techniques can encode biological information in a single structure that allows users to operate relationships, extract information, learn connections, and make predictions to discover potential new relationships between existing drugs and vector-borne diseases. In this article, we compared seven knowledge graph embedding models (TransE, TransR, TransH, UM, DistMult, RESCAL, and ERMLP) applied to Drug Repurposing Knowledge Graph (DRKG), analyzing their predictive performance over seven different vector-borne diseases (dengue, chagas, malaria, yellow fever, leishmaniasis, filariasis, and schistosomiasis), measuring their embedding quality and external performance against a ground-truth. Our analysis found that no single predictive model consistently outperformed all others across all diseases and proposed different strategies to improve predictive performance.