FAKE NEWS DETECTION USING PYTHON AND MACHINE

LEARNING HARINI. T*, ARUNA. S**, LAVANYA .K***

(Students of 3rd year B.Tech (IT) PANIMALAR ENGINEERING COLLEGE, CHENNAI)

ABSTRACT

Social media fake news detection is a novel field that is developing right now. Currently, the society
is significantly impacted by social media news, as evidenced by the statistics of people using
Facebook, Twitter and other social media platforms. Use apps like WhatsApp to share the most
recent news whether it is true or false. More information is being produced and shared by consumers
than ever thanks to the widespread use of social media platforms, many of which are false and have
no bearing on reality. It is suggested in this paper to classify news articles automatically using an
ensemble machine learning method. It aims to provide the user who has the ability to judge whether a
news item is accurate or not and to verify the reliability of the website that is posting it.

KEYWORDS: Internet, Social Media, Fake news.

I. INTRODUCTION presidential election was even further prevalent

of Facebook than the most widely circulated
More and more people are choosing to search actual conventional news. Consumers are
for and consume news from social media rather purposefully led to believe inaccurate or
than conventional news sources as we spend an prejudiced information by fake news. Fake news
increasing amount of our lives interacting modifies how people view and react to real
online through online media platforms. It news. This research aims to compare the
became simpler for customers to get the most effectiveness of various algorithms at identifying
recent news at their hands thanks to social fake news.
media websites. These platforms are, however
also utilised negatively to slant perceptions an There are two different types of algorithms: the
manipulate attitudes. The occurrence is first type uses a manually labelled news dataset,
frequently referred to as false news. It is simple and the second type supports the concepts of
to share and debate news with friends, and its using AI in conjunction with a manually labelled
frequently less costly and more timely to dataset to identify fake news. There two articles
consume news there. on related subjects vary in that the latter one
used Logistic Regression, specifically for the
For instance, 62% of American adults over the purpose of identifying bogus news, and that the
age of 18 in 2016 accessed news online. While current data collection was used to evaluate the
in 2012, only 49% said they viewed news on the built algorithm, which gives the chance to
internet. However, due to low cost of online evaluate its effectiveness using neoteric
news publication and because it spreads through statistics.
social media much more quickly and easily. It is
obvious that the most commonly disseminated
false information during the 2016 United States II. LITERATURE SURVEY
This study examines the various methods or context information into account.
systems that have been employed in the past to
identify fake news. This papers main goal is to Feng, Banerjee et.al.,[2]are successful in
observe and identify the most effective and deception-related classification tasks utilizing
objective solutions to the given issue. online review corpora, achieving 85%-91%.
Additionally, the survey below examines each These methods have only been proven successful
methodology used in the literatures discussed when combined with more advanced analytical
[3]. Fake news has puzzling root causes and is methods that use deep syntax analysis.[5]
widely spread. Numerous strategies are Top 5 Top 5
available and have been adopted by both Untrustworthy Trustworthy
people and organisations. News Sources News Sources

Regardless of the approaches, tools and Before It’s 2068 Wall Street 3898
resources used, this process is more or less News Journal
followed in other surveyed literatures. As a
result, it can be seen that machine learning is a Zero Hedge 146 New York 836
popular field for text analysis. It appears that a Times
false news detector is an unofficially named Guardian 90 USA Today 824
data science implementation model that can
identify and categorise fake and real news Washington 79 Washingt 823
based on provided data. [8] Since binary Examiners on Posts
classification is the focus of the news detection
problem, machine learning methods like IV. PROPOSED SYSTEM
logistic regression, Supported Vector Machines
(SVM), and Naive bayes are used more It may be useful to utilise a tfidf matrix, or word
frequently. tallies based on how frequently they appear in
other articles in the given dataset. This work
III. EXISTING SYSTEM develops a model using the count vectorizer.
Building a Naive Bayes classifier will be ideal
The classification of online reviews and because it is common for text based processing
publicly accessible social media posts has been and this challenge involves text categorization.
the focus of the majority of the research on The real objective is deciding which type of text
machine learning algorithms for fraud transformation (count vectorizer vs tfidf
detection. In the literature, the problem of vectorizer) (headlines vs full text). The next step
spotting “fake news” has drawn a lot of is to extract the best traits for the count
attention, especially since late 2016 during the vectorizer or tfidf vectorizer. To do this, a large
American Presidential Election. number of the most widely used words and or
phrases, whether they are capitalised or not, are
A number of strategies are described by used, and most stop words are largely removed.
Conroy et. al.[1] with the purpose of accurately In addition to this, Power BI is used to visualize
classifying the deceptive articles. They point the dataset in graphical representation.
out the superficial parts of speech(POS)
tagging and V. NAIVE BAYES CLASSIFIER AND
simple content-related n-grams have typically ITS USES
been inefficient for the classification challenge
because they neglected to take important Naive Bayes classifiers are a type of
straightforward machine learning used in
artificial intelligence.The well-known Naive
Bayes approach employs multinomial NB and
pipelining concepts to assess the accuracy and
veracity of news. There are several algorithms
for training these classifiers that focus on
common concepts, thus it is not the only one.
You can use Naive Bayes to determine whether
the news is authentic or bogus.

VI. NAIVE BAYESIAN FORMULA

DETAILS

The recipe of being naive is as follows: The

likelihood of the prior occurrence is used in
Bayes classification, which contrasts it with the
current event. Every single after calculating the
event’s probability, it is then determined how
likely the news is overall given the dataset. By
calculating the overall likelihood, we may
therefore obtain an approximation of the value
and determine whether the news is accurate or
not.

P(C / D)=P(D/ C). P(C )/ P (D),

Finding the probability of event, C when event

D is TRUE
VII. METHODOLOGY
P(C) = PRIOR PROBABILITY In this paper it provides an explanation of the
three-part method. The main component uses a
P(C / D) = POSTERIOR PROBABILITY machine learning classifier and is inactive. We
FINDING PROBABILITY: looked at the model and trained it with four
various classifiers before selecting the best one
P(C / D1)=P(C 1/D 1). P(C 2 /D 1) . P(C 3/ D 1) to use in the end. The second component is
dynamic and uses the user’s keyword or text to
P(C / D2)=P(C 1/D 2). P(C 2 /D 2). P(C 3 /D 2) look online for information about the likelihood
that the news is true. The third section provides
P(Word) = Word count + 1/(total number of
proof that the customer actually contributed the
words + No. of unique words) if probability is
0. Consequently, one can determine the news URL.
accuracy by applying this method

A.System Design
 of categorizing the domain if the location isn’t
included in either database, the implementation
merely states that the news aggregator dosen’t
exist.

VIII. RESULT:
A Python programming tool was used to interpret
B. System Architecture the results for specific data sets. Results are
presented in various tables and histograms.
i) Static search
Table 1 Dataset evaluation result
The design of the static part of the false news
Outcomes estimate
detection system is rather simple, and it is
finished by keeping in mind the key AI Correctness 95.26814
measure stream. The frameworks’s
configuration is self-explanatory and is given Fidelity 95.79288
below. Most of the steps in the design are
ii) Dynamic search Rescinding 94.56869

F-measure 95.17685

The evaluation findings for a particular data set

are displayed in Table 1. The accuracy of the
model dataset was 95.26 percent, the accuracy
of the outcome was 95.79 percent, and the
accuracy of the Recall and F-Score were,
respectively, 94.56% and 95.71%.

Table 2 class results predicted are

Model Prediction Predicti
Yes on No
The websites’s second search box asks for
particular keywords to be entered for web Real Yes 296 17
searches and displays the possibility that those
phrases will actually occur in a piece or in the Class No 13 308
same post that makes use of those keywords.
With this model, we have 296 positive traits,
iii) URL Search 308 negative attributes, 17 false positives, and
13 false negatives, as shown in Table 2 of the
The execution searches for placement in our projected class results..[6]
data set of actual locations or boycotted
locations database after receiving a particular
Table 3 Actual news and Fake news
site name in the third search field of the Table 3 shows predictions for true and false
positioning.the domain names that routinely news, with True Positives(TP) and False
provides accurate and reliable news are kept in Positives(FP), respectively.[6].
the verity sites database and vice versa. Instead
 Predicted verify the authenticity of websites.The accuracy
Model
Class of the dynamic system is 93% and gets better
with each repetition.
Real Yes TP FN
References:
Class No FP TN
[1] N. J. Conroy, V. L. Rubin, and Y. Chen,
“Automatic deception detection: Methods for
finding fake news,” Proceedings of the
Association for Information Science and
Technology, vol. 52, no. 1, pp. 1–4, 2015.

[2] S. Feng, R. Banerjee, and Y. Choi, “Syntactic

stylometry for deception detection,” in
Proceedings of the 50th Annual Meeting of the
Association for Computational Linguistics:
Short Papers-Volume 2, Association for
Computational Linguistics, 2012, pp. 171–175.

[3] International Journal of Computer Science &

Communication (ISSN: 0973-7391) Volume 12
This diagram depicts the range of fake and real
Issue 2 pp.38-44 April 2021 - Sept 2021
news detected in the given dataset.
www.csjournals.com
IX.Conclusion:
[4] International Journal of Recent Technology
The fake news detection system recommends and Engineering (IJRTE) ISSN: 2277-3878,
using the user’s input, which is then qualified as Volume-8, Issue- IC2, May 2019
true or false. Various NLP and machine learning
techniques should be applied in this case. A
[5] Learn in Fake news detection using machine
suitable dataset should be used to prepare the
learning in Pantech solutions.
model, and several performance measures
should be used to complete the performance
[6] Fake and Real News detection Using Python
evaluation. The best model, the most accurate
International Journal of Scientific Research in
models, was used to aggregate headlines or news
Science and Technology . june 2020 DOI:
articles. The best Logistic regression model was
10.32628/IJSRST207376
obtained from the static search with 65%
accuracy. Therefore, Logistic regression
[7]Learn on https://medium.com/swlh/fake-
performs better with search parameter
news-detection-u sing-machine-learning-
optimisation achieving 75% accuracy.
69ff9050351f
Thus, this demonstrates that there is a
[8] Detecting Fake News using Machine
75%chance that a customer’s real sentence will
Learning: A Systematic Literature Review
be represented by a certain news article or
headline that they enter into our model. Users
can view news articles or keywords online; also
[9] Fake News Detection Using Machine
Learning P. Yogendra Prasad1 ,
Dr.G.Nagalakshmi2 , P. Siva Kumar3 1Assistant
Professor, Dept. of CSSE, Sree Vidyanikethan
Engineering College, Tirupati. 2Assistant
Professor, Dept. of Computer Science, National
Sanskrit University, Tirupati. 3Applications
Lead, Oracle Corporation, Bangalore.

[10] Iftikhar Ahmad, Muhammad Yousaf, Suhail

Yousaf, Muhammad Ovais Ahmad, “Fake News
Detection Using Machine Learning Ensemble
Methods”, Complexity, vol. 2020, Article ID
8885861, 11 pages, 2020.
https://doi.org/10.1155/2020/8885861

[11] Smitha N, Bharath R (2020) Performance

comparison of machine learning classifiers for
fake news detection. In: 2020 Second
international conference on inventive research in
computing applications (ICIRCA), pp 696700.
IEEE, Coimbatore, India (2020).
https://doi.org/10.1109/ICIRCA48905.2020.918
3072

[12] Granik M, Mesyura V (2017) Fake news

detection using naive Bayes classifier. In: 2017
IEEE 1st Ukraine conference on electrical and
computer engineering UKRCON, pp 900903.
https://doi.orf/10.1109/UKRCON.2017.8100379