Global Journal of Computer Science and Technology, C: Software & Data Engineering, Volume 22 Issue 2

Hence, people started consuming alcohol in large numbers. Like this news, rumors were being spread across the world and people started to believe that news. In this work, we have presented a new data set to detect fake news. For this data set, we have gathered more than three thousand news from various news sites. In our data set, there are two types of news, one is true news another one is fake news. There are one hundred and seventy fake news in our data set. For detecting fake news from our data set, we have used four classification algorithm they are support vector machine and passive aggressive classifier. The rest of the work is organized as follows: in chapter II, we have reviewed some of the previous work. In chapter III, we have discussed the algorithm that we have used for our work. In chapter IV, we have discussed our work. In chapter V, we have discussed the performance and have concluded our work in chapter VI. II. L iterature R eview Previously many researchers have worked on this particular area. Some of them created new models or explore new areas to detect fake news. Others improved the existing model to improve fake news detection and others build a robust data set to fuel the detection system. In this section, we explore some of the previous works. Ruchansky et al. presented a fake news detection System using a Hybrid Deep Model. They proposed a model called the CSI. Capture, Score, and Integrate are the three modules of CSI model. The capture module is based on text and reaction. This module captured the engagement between users and news articles, the second module score learns the source feature. Then they combined two modules and integrated them with the third module to produce a label for fake news [29]. Kai Shu et al. presented a brief analysis of detecting fake news from social media, they also discussed fake News Classification based on social context and psychology, also reviewed current algorithm in terms of data mining [32]. Lutzke et al. conducted a 3 by 2 experimental model having guidelines, enhanced guidelines, and a controlled part. Each experiment has two types of news about climate change one is fake and another is real. A total of 2750 people from different fields participated in their experiment. Each participant was randomly allocate to one of six possible experimental variations [22]. Bahad et al. used deep learning model, Bi-directional Long short- term memory (LSTM) over other techniques like as Convolutional Neural Networks (CNN), unidirectional LSTM, Recurrent neural network (RNN), for detecting fake news [6]. Abdullah et al. used machine learning III. A lgorithms Previously many algorithms have been developed to detect fake news. We have used four different types of supervised learning algorithms to detect fake news. We briefly discuss the four existing detection algorithms in this section. Fake News Detection: Covid-19 Perspective Global Journal of Computer Science and Technology Volume XXII Issue II Version I 2 Year 2022 ( ) C © 2022 Global Journals In machine learning, Supervised learning is the task of creating a function that maps an input to an output based on given input-output pairs [30]. In supervised learning, the program is given labeled input data and the output result is expected. In regression and classification problems, supervised learning works fine, such as deciding what group a news story belongs to. and deep learning technique to detect fake news from Twitter data set. They have used five different Machine Learning algorithms, like Support Vector Machine, algorithm, Logistic Regression, and Recurrent Neural Network to detect fake news from the data set [1]. Zhang et al. used a two-layer method that includes the identification of fake topics and fake incidents [38]. Yang et al. detect satirical news using linguistic features [37]. Apuke and Omar developed a complete model studying from uses and gratification view [14]. Agarwal et al. used the natural language processing and the machine learning method to detect fake news [2]. Granik and Mesyura used a Bayes classifier to classify news from BuzzFeed data sets [17]. Yang presented Liar data set containing news more than 12 thousand for fake news detection. The data set is divided into six classes [36]. Fake News Corpus by Szpakowski. He uses multiple corpora to develop and test various models. The corpus has been automatically crawled using open sources.co labels [33]. Khan et al. performed benchmark research to compare the performance of several machine learning methods on three distinct data sets [21]. Zhang and Ghorbani gave a detailed review of what has been discovered so far on false news. They have described the detrimental impact of online fake news as well as the current status of detecting tools [39]. Alkhodair et al. presented a novel method for automatically identifying rumors, combining the learning of word embedding with the training of a recurrent neural network with two separate goals [3]. Bayes Naïve - Naïve- a) ”probabilistic classifiers” which is build from applied assumptions between the attributes. Due to its simplicity, speed, and good accuracy Multinomial classifier for text classification. Multinomial classifier learns a conditional probability that the nth document form a x n bayes classifiers are members of bayes theorem with strong ( ) independent bayes [23] is a common Multinomial Naïve Bayes Naïve Naïve Naïve multinomial bayes, logistic regeression, Naïve