LITERATURE REVIEW
In this study, Hakak et al. identify the most significant
characteristics that influence fake news classification. Ensemble models
are used to achieve optimal accuracy in classifying fake news datasets.
The ensemble classifier requires less training time. From the fake news
datasets, the proposed model extracts significant features and then
classifies the extracted features using a hybrid ensemble model
consisting of Decision Tree, Random Forest and Extra Tree Classifier. A
training accuracy of 99.8% and a testing accuracy of 44.15 percent were
achieved on the Liar dataset. Training and testing accuracy were both
100% in the ISOT dataset. In its paper, Hossain et al. discuss the fact
that feature engineering can be used to effectively address this issue:
limiting the spread of fake news at the source, not after it has become
a global problem. We manipulated text with extracted features, resulting
in an effective level of fake news detection.
A deep learning model was then developed and tested. Using this study,
it was successfully shown that the original features had an impact on
deep learning models that were unknown.
By analyzing the accuracy of a report and predicting its authenticity,
we propose a model for detecting fake news (Agarwal & Dixit, 2020). By
extracting features from the textual information and constructing
credibility scores, this model builds an ensemble network that can
simultaneously learn the depictions of news reports, authors, and
titles. A variety of machine learning algorithms are used for higher
accuracy, including SVM, CNN, LSTM, KNN, and Naive Bayes, and the LSTM
algorithm shows the best accuracy at 97%. Using precision, recall, and
the F1-Score, we evaluated the performance and effectiveness of
classifiers. Different algorithms were used to show the effectiveness of
the performance. Kaliyar et al., 2019), In this work, the author
proposes an ensemble machine learning framework based on a tree-based
gradient boosting technique, which combines content characteristics and
contextual features for the detection of fake news. Recently, gradient
descent algorithms have been derived as adaptive boosting methods for
classification problems. A single objective function is optimized using
this formulation, which shows why specific elements and parameters are
used in the methods. We apply various machine learning models for
classification based on a multi-class dataset (FNC). Comparing the
ensemble framework to existing benchmark results, experimental results
demonstrate its effectiveness. For multi-class classification of fake
news with four classes, we achieved an accuracy of 86% by using
Gradient Boosting algorithm (an ensemble machine learning framework).
(Elyassami et al., 2022) classifies news as fake or real by using
machine learning models. A total of five classifiers were developed
using Random Forest, Support Vector Machine, Gradient Boosting, Logistic
Regression, and Naive Bayes. Open-source datasets extracted from online
sources covering a variety of domains were used to train the models.
Using text lemmatization, vectorization, and tokenization, valuable
information was extracted from news text, increasing generalization and
accuracy of fake news classification models. An investigation of how
voting strategy impacts ensemble learning models was conducted. The four
performance measures used to evaluate the five classifiers were
accuracy, F1-score, recall, and precision. We are encouraged by the
results. It is possible to use these ensembles against fake news
spreading since they outperform other classifiers when trained with
random forest algorithms and gradient boosting algorithms. In this
paper, (Masciari et al). present their complete framework for detecting
fake news and describe their machine-learning-based solution. Using two
well-known and widely used real-world datasets, we demonstrate that our
settings are superior to state-of-the-art algorithms and are capable of
detecting fake news with high accuracy even in the absence of complete
content information.