This chapter explores deep learning methods for polymer property prediction, focusing on classification and regression tasks. It examines neural network architectures, including RNNs, LSTMs, GRUs, Transformers, and GNNs like GCNs, GraphSAGE, GATs, and GINs, for processing polymer data as sequences (e.g., SMILES) or graphs. The chapter discusses dataset preparation, evaluation metrics, and specific techniques like hyperparameter tuning and data augmentation to enhance model performance. Additionally, advanced learning paradigms, including transfer learning, multi-task learning, and self-supervised learning, are introduced to improve predictions with limited labeled data. These methods enable the accurate prediction of polymer properties, supporting innovations in polymer informatics.