Traditional recommendation algorithms often fail to capture the temporal dynamics and interaction sequences critical for accurately modeling user behavior. User interactions are typically influenced by the sequence of prior choices, necessitating a model that can effectively model such dependencies. This study introduces a novel approach that integrates the inverted Transformer model to enhance predictive performance in recommendation systems. The iTransformer, tailored to handle sequences of user-item interactions over time, utilizes an attention mechanism that treats variates as spatial dimensions, enabling the model to capture intricate dependencies between user behavior and item features across different time points. The model was evaluated on the real-world MovieLens dataset, demonstrating significant improvements over existing transformer-based models. Source code is available at GitHub .