Social media is one of the very powerful media in spreading information. People are interested in sharing without any proper checking of any sort of false information. Unstructured text data may be classified into meaningful categorical classifications using text classification, which is a typical study area in the discipline of Natural Language Processing (NLP). The main contribution of this article is to identify a finest framework to tackle the fake news problem with the NLP and Machine Learning techniques. In this empirical research, the fake news data is analysed with the different combinations of Vectorizers and Machine Learning Classifiers. From the experimental results on five benchmark datasets namely fake_real_news dataset extracted from Kaggle, COVID-19 Constrain, Politifact, ISOT and Gossipcop, it is observed that the fake news detection with the combination of TF-IDF Vectorizer and Passive-Aggressive Classifier outperforms the other existing methods.