Fake news is on the rise on social media, with fake news are influencing attitudes and decisions on any matter to have a serious impact on social coexistence. The base of fake news focus on content that excites and stimulates the consumer's emotions, so it spreads more easily and faster than real news. Verification the informed source of the news to determine whether it is real or fake news takes time to find. Using a model to predict fake news from social media through data mining techniques is a way to help prevent the spread of fake news. Also can be used to stimulate the dissemination of news. This research focuses on the essential properties of fake news and compare the prediction models of fake news with data mining techniques, as well as find the correctness of classification by using perceptron algorithm of Neuron Network, Decision tree method, the K-Nearest Neighbors and Naïve Bays method. The measurement using the accuracy, the mean absolute error (MAE) and the mean square error (MSE). The results from the efficiency of predicting, the method of classifying information of four algorithms found signifies the competitive accuracy degrees of neuron network is 95.78%, the MAE is 0.2011 and the MSE is 0.1915. However, the second proficiently accuracy, MAE and MSE of the K-Nearest Neighbors is 90.51%, 0.2051, 0.2315 respectively. This research could be used as a prototype for the further construction of an automated fake news detection system.