A rapid rise and out-reach of social data and the web have led to the broadcasting of dubious and untrusted content a wider audience, which is negatively impact on people’s life. This research study focuses on fake and original news classification based on features and unseen patterns. Over the past decades, many of research studies have been conducted to tackle the detection and identification of fake news. In this paper, we focus on classifying the fake news using different machine learning algorithms such as LSVM, Perceptron, KNN, Random Forest (RF), KNN and so on. The actual challenge is the lack of an efficient way to tell the difference between real view and fake on, even sometimes humans are also confused and can’t differentiate. The proposed system works on two steps 1) get the relevant article data and match with the knowledge database and secondly, it identifies the patterns and underlying style of fake content. The classifier has ability to detect the fake news on newly introduced fake news dataset. The experimental result shows that the given classification model obtains up to (96 %) accuracy on Decision tree, AdaBoost and LR approach as compared to other machine learning algorithms.