DETECTING AI-GENERATED FAKE NEWS USING MACHINE LEARNING

Abstract

The spread of AI-generated fake news poses increasing obstacles to the distribution of information as sophisticated large language models like GPT continue to evolve. Due to their limited ability to discern between real and fake news, traditional text classification techniques are unable to identify such content. This paper presents an MLP (Multi-Layer Perceptron) Classifier combined with Natural Language Processing (NLP) methods for identifying fake news produced by artificial intelligence (AI) in order to address this problem. Tokenisation, stopword elimination, and vectorisation are preprocessing techniques used to extract significant features from textual data, which are subsequently fed into the MLP network. The classifier captures intricate linguistic characteristics that define fake news by utilising nonlinear activation functions and several hidden layers. To train and assess the system, a fresh dataset comprising 42 news categories was created using GPT-4. According on experimental data, the suggested MLP model outperforms conventional machine learning techniques by achieving strong F1 scores and dependable accuracy. These results demonstrate how MLP-based designs can improve the detection of fake news and protect the integrity of online information.