One common post-transcriptional modification that is essential to biological activities is RNA 5-methylcytosine (m5C). A large amount of RNA data containing m5C modification sites has been gathered as a result of the rapid development of high-throughput sequencing technology. While there are a lot of machine learning based techniques available for identifying m5C alteration sites, these models' accuracy still has to be raised. This study proposed a novel method, Feadm5C, which predicts m5C based on fusing molecular graph features and sequencing information together. 10-fold cross-validation was used to assess the model's predictive performance. In addition, we used t-SNE visualization to assess the model's stability and effectiveness. While keeping feature encoding and model structure straightforward, the approach suggested in this work outperforms the most recent approaches in use. The dataset and code of the model can be downloaded from GitHub (https://github.com/LiangYu-Xidian/Feadm5C).