A reliable fake news detection model using a hybrid convolution (1D-2D)-based adaptive temporal convolutional network with feature extraction

News is considered the essential element in an individual's day-to-day life to update their knowledge of the world. Yet, the news websites are circulating minimal quality news with more misinformation, and this can easily spread as fake news among individuals. Fake News Detection (FND) framework has been introduced recently to curb the fake news transmission on social media, and websites, but the system requires manual intervention. Yet, it is a rigorous task to manually identify fake news among large media platforms. Hence, it is necessary to identify Fake news through advanced automated deep learning models. The social media, news websites and several sources provide multimedia data in large volumes. Thus, a novel FND model by deep learning is proposed in this work to automatically identify Fake news from real information. At first, the news containing various text and image files is collected from the benchmark dataset sites. For extracting the features from text data, the Text Convolutional Neural Networks (TextCNN) method is used, and it identifies the text-related features as feature set 1. The image is directly given for learning and it is represented as feature set 2. Next, a Hybrid Convolution (1D-2D)-based Adaptive Temporal Convolutional Network (HC-TCN) is proposed in this work for processing high-resolution image features set 2, and the text feature set 1. The proposed HC-TCN processes both the information to define the class information related to Fake news and original news. For improving the effectiveness and accuracy of detection, the proposed HC-TCN is improved with the proposed Improved Lotus Effect Optimization Algorithm (ILEA). The final detection results of the implemented HC-TCN model are compared with other techniques to justify its robustness in FND detection.