Evaluation of Machine Learning Algorithms for Fake News Detection Based on Accuracy, Using Word2vec as Feature Extraction Method
Aminu Malam Ya’u , Abubakar Muhammad Miyim
Keywords: Machine learning, fake news, word2vec, MLP, SVM.
Fake news refers to false information presented as news, circulated over the internet or other communication media for either political purposes or joking. Fake news is a serious problem that needs to be addressed because it causes a lot of issues in human lives such as political crises, fraud, murder and so ‘on, to resolve such problems, there is a need to develop a machine learning model that can be classified whether the news is fake or real. Much research works on fake news detection has been carried out, literature review shows that researchers employed many machine-learning techniques for the detection of fake news. In this research, the word2vec feature extraction technique was adopted in converting the news text into numerical values so that it can be fed into machine learning models. Four Machine learning (ML) classifiers were selected, trained, and evaluated to choose the best classifiers for fake news detection. An experimental result shows that Multi-Layer Perceptron (MLP) outperformed the other algorithms with accuracy-98%.