An Improved Technique to Identify Fake News on Social Media Network using Supervised Machine Learning Concepts

Abstract

This paper presents an improved social media news separation system called unstructured Fake News Detection (UFND) system and it aims to identify the unstructured social media news data that belongs into fake or real class based on probability and improved naive bayes techniques. The proposed UFND consists of four phases likely pre-processing, training, matching and validation respectively. The technique identifies matching phrases over each individual data element based on predetermined key words model and ignore the irrelevant words in the respective document. In the later phase, the proposed system has train the pre-processed data set through the process of separating the data set into two classes namely fake and real based on probability technique. Further the system has identified the given test news document belongs to existing class label over the training data set based on improved Naive Bayes technique. The UFND system evaluates the performance over the result of previous stage. The UFNDS experimental results have demonstrated an outperforming results in segregating and identifying the fake news pattern over the unstructured social media news data set with good accuracy based on supervised probability methods.