Movie recommendation algorithms play an important role in assisting consumers in identifying films that match their likes. Deep Learning, particularly Long Short-Term Memory (LSTM) networks, has shown substantial promise in collecting sequential patterns to improve movie recommendations among the different techniques used for this purpose. Long Short-Term Memory-Inter Intra-metapath Aggregation (LSTM-IIMA) in movie recommendation systems is proposed in this study, with a specific focus on incorporating intra and inter-metapath analysis. The intra-metapath analysis investigates interactions within a single metapath, whereas the inter-metapath analysis investigates links between numerous metapaths. Intra and inter-metapath analyses are used in the LSTM-based movie recommendation system LSTM-IIMA to capitalise on these rich linkages. Each metapath sequence records the dependencies of a user’s interactions with films and other things. The LSTM architecture has been modified to handle these metapath sequences, processing them to record temporal dependencies and entity interactions. To optimize the parameters and minimize prediction errors, the model is trained using supervised learning techniques. To measure the quality and usefulness of the recommendations, the LSTM-IIMA evaluation incorporates metrics such as precision, recall, ablation analysis, time efficiency and Area Under the Curve (AUC). The performance of the system is compared to that of alternative recommendation techniques HAN and MAGNN. Overall, incorporating intra and inter-metapath analysis into the LSTM-IIMA improves its ability to capture complex linkages and dependencies between movies, users, and other things.