Analysis of Fake News In Social Medias for Four Months during Lockdown in COVID-19

All over the world, development of micro blog and other social platform indicate that Social Media is now the focus and trend of the Internet. Daily life, study and work are influenced by news in Social Medias . Micro blog is new emergent type of media and it spreads information rapidly in the crowd in recent years. Suppose an user searches for specific information about one topic on micro blog. He/she found easily plenty of information related to his/her search in social medias. The problem is to find out the correct information. Normally, multi-document summarization method deals with a collection of documents about one topic for extracting the valuable points and discards useless information. Actually, it needs to extract the topic content by adding topic factors and social patterns. Topic factor is the lexical information related to the topic. Social pattern relates to special interactive mode owned by online social network, such as comment and repost. People has been seen the fake news on mobile/internet during lockdown period. It is of no doubt that anyone with a social media account has seen at least one example of this.Humanity’s greatest challenges are to detect false information. Fake news are collected from 150 persons using social media The aim of the paper is to investigate the truthfulness of the news people share on social media using K-nearest Neighbour (KNN) based Classifier method.


Introduction
Fake news , if people believe, related to COVID-19 increases anxiety, stress, and even depression. The reason behind this is to stay-at-home, curfews, and closing of essential businesses. Many families are now phasing unemployment scenario. Some are forced to work from home due to the direction of companies. Many companies and businesses are working with loss due to coronavirus. This new life is stressful enough. So anxiety of people is clearly increased due to new life style [1].
Fake news is primarily spread via social media platforms, such as Facebook, Twitter, etc. Fake news is hard for some people to identify and can create confusion about true fact. People has doubt about accurate information. For example, When consumers learn information via fake news that the particular product is long-lasting. They have curiosity to see and purchase the product. Since it is not possible to purchase during lockdown or the information is incorrect. It made distress on consumers.
Researches have been going on to study the way fake news can be identified and detected [2].
It is true that the main way to combat fake news is by educating ourselves to understand fake news. Machine learning method may help to get identify false information during COVID-19.
Facebook and Twitter are now taking measures to remove fake news and misinformation from their platforms [3]. The main objective of the study is to make sure that the information received from social media/mobile/internet are reliable and true. COVID-19 pandemic in 2020 dominates the media, both domestically and abroad. Alongside it is required for the attention on the pandemic first as well as to stop viral spread of fake news online related to coronavirus.
At the time of COVID-19, the spread of fake news offers unique challenges and dangers to the public.
Machine learning approaches are applied for classification and recognizing fake news related to COVID-19. This paper requested 150 selected persons to address the problem of identifying fake news, people are requested to notify fake information in their social accounts. News in Mobile and email are also checked to get false news. Then machine learning based classifier is applied to analyse these fake news for measuring accuracy.

Related Works
Fake news is one of the serious issues in recent times addressed in the domain of COVID-19.
People often post fake news in social media to mislead people regarding shopping of product with less price. It may guide other people to visit shopping mall to see the product. These news can be manipulated and hence it may mislead other people to visit the place. Natural Language Processing (NLP) has been used to detect fake news by extracting features [4][5]. Next, machine learning techniques are applied on these features. Lexicon based approaches may be one alternative to machine learning techniques that use dictionary or corpus to eliminate false news.
Fake news in social media also hacked user accounts unknowingly. The fundamental study of fake news detection relies on three perspectives-how fake news is written, how fake news spreads and how a user is impressed by the fake news. Features related to news content and social context are extracted and a machine learning model is imposed to recognize fake news [6][7].
People share false claims about COVID-19 in part because they simply fail to think sufficiently about whether or not content is accurate. The news need to share with others without testing the news are correct or not. It can be of two types. Some participants are far worse at discerning between true and false content when they share on social media without knowing about accuracy of the fact. Participants engaged in more analytic thinking and had knowledge to separate fake news from the true news. They may sometimes discard the news and debarred the news spreading to social media. In this case, common sense is applied at the beginning of the study -i.e., asking people to judge the accuracy of a non-COVID-19-related news. It helps to stop spreading of false news in the social media. Some researchers found that these people can analyse false versus true news [4].

Data Set
The dataset contain news of 150 people obtaining from their social media accounts. These news are noted for analysis. The dataset are used in the proposed method from 20 th March 2020 to 8 th June 2020 for testing the overall performance of the approach. Before fitting this data to classifier, pre-processing is made on this dataset based on similarity news in social medias of 150 people. Pre-processing techniques include removal of not related to the corona, discard any communal news, deleted incomplete news. This prepares the dataset to be transformed into true features This features are the input to classifier.

Proposed Method and Results
The K-NN classifier is used for the value k=5 considering four months. Initially the classifier is trained actual data and is tested entire dataset for the prediction purpose. The performance measure metrics such as Accuracy, F1-score , and Cohen-Kappa score are used for evaluating the prediction. Accuracy identifies the ratio of true predictions over the total number of instances considered. However, the accuracy may not be enough for evaluating model's performance since it does not consider wrong predicted cases. If a fake news is treated as a true one, it creates a significant problem. Hence, it is necessary to consider false positive and false negative cases that compensate to misclassification. For measuring this compensation, precision and recall is quite necessary to be considered. Precision identifies the ratio of correct positive results over the number of positive results predicted by the classifier. Recall denotes the number of correct positive results divided by the number of all relevant samples. F1-Score is a parameter that is concerned for both recall and precision and it is calculated as the harmonic mean of precision and recall. Apart from all these measure, Cohen-Kappa Score is also considered for evaluation of the perfectness of the method. It is a statistical based measure that finds out inter-rate agreement for qualitative items for classification problem. The results are shown from Table 1 to Table 5. From figure 1 to figure 5 shows the related graphs.