This project entitled "RUMOR DETECTION ON SOCIAL MEDIA: An Experimental Study" submitted by Joy Sarker, Falguni Malakar and Sajol Kumar Roy to the Department of Computer Science and Engineering, Daffodil International University has been accepted as satisfactory for the partial fulfillment of requirements to the degree of B.Sc. Department of Computer Science and Engineering Faculty of Natural Sciences and Information Technology Påskelilje International University. We hereby declare that this project has been carried out by us under the supervision of TAPASY RABEYA, Associate Professor, Department of CSE, Daffodil International University.
We are truly grateful and wish our profound indebtedness to TAPASY RABEYA, Associate Professor, Department of CSE, Daffodil International University, Dhaka. We would like to express our deepest gratitude to Dr.Touhid Bhuiyan, Professor and Head, Department of CSE, for his kind help in completing our project and also to other faculty members and staff of CSE Department of Daffodil International University. We would like to thank all our coursemates at Daffodil International University who participated in this discussion while completing the coursework.
Social media platforms have been used for information and news gathering, and they are highly profitable in many applications. Numerous attempts have been made to detect and expose rumors on social media by analyzing its content and social context using machine learning procedures. These young people try to keep themselves connected by using social media. They share all their feelings of cherish and contempt, hostility and cruelty through these social media.
We are trying to capture the mood and behavior of young people in the stages of social media and trying to clarify social media as an uncontrolled and wild scene for young people.
Motivation
Rationale of the Study
Research Questions
Expected Output
Report Layout
Background ..................................................................... 3-7
Related Works ............................................................................................. 3-5
Many considerations centered on the recognition of the source of The Rumors and the recognition of the rumours. In this paper, we depict the challenges involved in fake news detection along with related messages. The sources of the dataset used in this study are briefly mentioned in the following sections.
After that part is the text of the news and the link searches the internet for the likelihood that the news is true. The static design element of the fake news detection system is quite straightforward and is built around the simple process of machine learning. The reason why random forest display works so well could be: A large number of moderately unrelated models (trees) working as a committee will beat any of the component person models.
Mathematically, the log odds of the outcome are demonstrated as an equal combination of the indicator factors. Its purpose is to make upgrades that remedy the crash, which causes unusually small changes within the standard of the weight vector. In subsequent times, the fake news area has concluded one of the widespread subjects among examiners all over the world.
Fake news, not because it was, remains the word as a few times, as it has recently become one of the biggest unbiased dangers. The development of internet use and social media use is increasing the advancement of off-base news. Using this formula, we get a text accuracy from the rumor detector of 71.8%.
Using this formula we get the accuracy of the text of rumor detector is 75.9%. 34;Rumor detection on social media with bidirectional graph convolution networks." In Proceedings of the AAAI Conference on Artificial Intelligence, vol. 34;Leveraging the implicit structure within social media for emergent rumor detection." In Proceedings of the 25th ACM international on conference on information and knowledge management, pp.
Scope of the Problem
Challenges
Research Methodology ................................................ 8-17
Research Subject
Our research topic is finding rumors about a text or link to see how people generally express the rumor while writing.
Data Collection Procedure ........................................................................ 9-10
10 Figure 3.1 shows the collection of our raw data, which we have collected from various sites like Facebook, newspaper, website, etc. The third and last section verifies that the URL entered by the user is valid. Python comes with a large number of libraries and extensions that can be easily used in Machine Learning.
Almost all forms of machine learning algorithms are available for Python allowing easy and fast evaluation of Machine Learning methods (Flask Web App). The site's second search form requires specific words to be searched online and gives a suitable % certainty of that term being included in a paper or piece comparable to those word references. The third field of site view accepts a particular title of the site space and use looks our true destinations or boycotted location databases for the location.
To achieve more better encounters, it is crucial to clear the data a few times as it can be used to illustrate in advance. Cleaning the fabric data is essential to highlight features that we pointed out required our machine learning system to pick up on. But for our vectorizer that counts the number of words and not the setting, this doesn't count, so we evacuate all unusual characters.
In particular, we investigated four different machine learning calculations Multinomal Naive Bayes Detached Forceful Classifier and Calculated regression. This method is based on the Bayesian hypothesis, which recognizes that the proximity of a particular focus in a course is independent of the proximity of any other inclusion. To classify modern protest based on quality, each tree gives a classification, meaning it defines a class.
It deals with firing and includes arbitrariness when building each person tree to undertake to make an uncorrelated forest of trees whose forecast by committee is more accurate than that of any person tree. Irregular Woodland, as the title suggests, consists of a large number of person-choice trees that act as an outfit. Each character tree in the irregular forest spits out a price expectation, and the path with the most votes becomes our model's expectation.
It is used to evaluate discrete values (parallel values such as 0/1, yes/no, real/fake) against a given set of free values. Such calculation remains passive for a correct classification result, and becomes powerful in case of miscounting, updating and modification.
Implementation step
Experimental Results and Discussion
Experimental Results......................................................................... 18-19
Summary, Conclusion, Recommendation and
- Impact of society
- Conclusions
- Recommendations
- Implication for Further Research…
- Future works…
To classify the expressions into either true or false using rumor detection to detect rumors, that was our main goal. For each text and link an expression will be generated whether it is true or false. After this experimental result, we finally came to a decision that the text or link is fake or real.
As a result, our project will be able to detect the news that is real or fake. It can make a big impact on social media and in our daily life. It was not so easy to detect rumors from a text or link, as people do not agree to identify exact rumor of the same sentence.
Our text and trained link-based data help us uncover the exact rumors most people are thinking about. Among the different approaches, we have used data to extract semantic information from a text or link to provide accurate gossip. So generating the correct expression from the given input text &link will fail if it does not contain specific lexicon.
The experimental study that we have carried out on the detection of rumors with a satisfactory result is leaving a strong mark on our work. Research can be done on how our project can be used to build something for disabled people especially those who are blind. We will work with a large amount of data so that the result is more accurate.
34;Rumor Detection Using Machine Learning Techniques in Social Media." In Chinese Natural Language Processing and Computing, pp. 34;Multiple Time Series Data Analysis for Social Media Rumor Detection." In 2018 IEEE International Conference on Big Data, p.