You are on page 1of 1

Fake News Detection System Using Machine Learning

Team ID: 99 Supervisor: Prof. Dr. Nilanjana Dutta Roy


Student Name: Pronoy Bosh Roll No.: 189
Student Name: Md Imran Hossain Roll No.: 190

Introduction Discussion
❑ As the social media and internet is rapidly adopted in ❑ These vectors are then used in finding out the
today’s world, some misleading informations are Cosine similarity between each sample. We ❑ By this project people can easily identify the fake
being created and shared almost everyday now. So, studied and trained the model with different news. The ensemble learner performed better in
in this project we propose to use a machine learning classifiers and chose the best classifier for final comparison to other learning models on all
ensemble approach for the automated classification performance metrics. Logistic regression is a relatively
execution.
of news articles as fake and real. simpler model but achieved an average accuracy of
over 90%.

Objectives and Challenges


❑ The primary aim of our research is to identify
patterns in text that differentiate fake articles from
Results Conclusion and Future Scope
true news. And by using this method people can
identify and expose the scammers.
❑ Fake news detection is a big challenge. Because
the social media users’ number is a very huge From tests and example, we can mention that ❑ In this research, we present the task of automatic
number now-a -days and it’s very difficult to identify overall, the best performing algorithm is detection of fake news using machine learning
the particular ones. bagging classifier (decision trees) (accuracy 94%), models and ensemble techniques. We have used
whereas the worst performing algorithm is some new publicly available fake news datasets. The
Wang-Bi-LSTM (accuracy 64.25%). Individual primary aim of the research is to identify patterns in
learners’ accuracy is 77.6% whereas the accuracy text that differentiate fake articles from true news.
Our Approach of ensemble learners is 92.25% in precision, recall
❑ By using those tools and the machine learning and score. References
algorithms such as Logistic Regression, Support
Vector Machine, K-Nearest Neighbors (KNN) etc., ● [1] A. D. Holan, 2016 Lie of the Year: Fake News, PolitiFact,
we can do the pre-processing on the datasets such Washington, DC, USA, 2016.
as all the NULL valued rows are dropped. The
textually based columns from the datasets are first ● https://www.engadget.com/2018/12/10/more-people-get-ne
n-gram converted and then into TFIDF vectors. ws-from-socialmedia-than-newspapers/

● A. Robb, “Anatomy of a fake news scandal,” Rolling Stone,


vol. 1301, pp. 28–33, 2017.

Fig. 2 ● https://scroll.in/latest/838693/fake-news-is-a-concern-for-83
-of-indianmediaconsumers-reveals-bbc-study

Fig. 1

You might also like