<p>The rapid digitization of healthcare systems has led to an unprecedented surge in healthcare data, necessitating robust and automated methods for data classification. In this study, an innovative approach to automated healthcare big data classification is presented by combining the gazelle optimization algorithm with multimodal deep learning techniques. The research problem addressed revolves around the need for accurate and efficient classification of diverse healthcare data sources, such as patient records, diagnostic measurements, and clinical notes. Automated classification is crucial for early disease detection, treatment recommendation, and resource allocation. However, this task is inherently challenging because healthcare data are diverse and it needs reliable classification models. The proposed methodology leverages multimodal deep learning models, including deep echo state networks for sequential records and deep fusion models for structured and unstructured text. This multimodal approach enables us to effectively capture and integrate information from diverse healthcare data sources. The DL technique improves the computational efficiency and prevents the vanishing gradient problem. To optimize the models and enhance classification accuracy, the gazelle optimization algorithm is introduced that dynamically adapts model configurations during training. Gazelle optimization algorithm (GOA) optimizes hyperparameters by mimicking the efficient and agile characteristics of the gazelle. This proposed approach yielded 98.57% accuracy, 97.02% recall, 96.89% precision, and 98.31% F1 score in the healthcare classification. The proposed approach paves the way for more accurate and efficient healthcare data classification, ultimately benefiting healthcare providers and patients.</p>