The remarkable success of machine learning models has sparked considerable interest in multimodal data fusion techniques. Addressing the challenge of integrating diverse data modalities while enhancing classification performance remains a key focus. In this paper, we introduce FAT-LSTM, a novel framework designed to fuse multimodal data for classification tasks. Leveraging advanced Gating and Attention-based Long Short-Term Memory (LSTM) mechanisms, FAT-LSTM represents a significant improvement in multimodal learning field. Through extensive comparative analysis against established baseline models, our study shows the superior performance achieved by FAT-LSTM. Additionally, thorough ablation analysis provides insights into the inner workings of the model, shedding light on its effectiveness. Empirical validation across multiple space weather datasets further confirms FAT-LSTM’s efficacy across various scenarios. This research highlights the crucial role of multimodal data fusion in effectively addressing real-world challenges.