Hierarchical multi-label academic text classification (HMTC) is to assign academic texts into a hierarchically structured labeling system. We propose an attention-based hierarchical multi-label classification algorithm of academic texts (AHMCA) by integrating features such as text, keywords, and hierarchical structure, the academic documents are classified into the most relevant categories. We utilize word2vec and BiLSTM to obtain embedding and latent vector representations of text, keywords, and hierarchies. We use hierarchical attention mechanism to capture the associations between keywords, label hierarchies, and text word vectors to generate hierarchical-specific document embedding vectors to replace the original text embeddings in HMCN-F. The experimental results on the academic text dataset demonstrate the effectiveness of the AHMCA algorithm.