Rumor classification is designed to classify fake information as fake or real. The significant goal of rumor detection is to reduce the rapid proliferation of false information. Fake news spreads in varied forms such as text, image, audio, and video. Most of the rumor classification models are solely focusing on fake textual information. The necessity of multi-modal approaches in rumor classification to handle the multiple features of misinformation. Multi-modal rumor classification is able to classify the multiple features of the same fake news.

   Traditional methods for multi-modal rumor classification are insufficient to produce an accurate classification. The ensemble learning model is the learning model that combines multiple learning models to solve a task of a single learning model and improve classification performance. Ensemble learning is well suitable for multi-modal rumor classification. Integrating deep learning with ensemble learning helps in automatically extracting high-level features using multiple deep learning models and formulating a prediction before combining the final prediction. Deep ensemble learning models produce better generalization performance for multi-modal rumor classification.