An Ensemble Learning Frame Work for Robust Fake News Detection

 

An Ensemble Learning Frame Work for Robust Fake News Detection

Neela K Assistant professor Information Technology Sri Sairam Engineering College Autonomous institutions Chennai

Deekshitha K UG Student Information Technology Sri Sairam Engineering College Autonomous institutions Chennai

Deepa R UG Student Information Technology Sri Sairam Engineering College Autonomous institutions Chennai

Abstract-

In today’s digital age, the widespread use of the internet has transformed how people access news, with mobile apps and social media platforms taking over traditional media channels. However, this shift has brought about a concerning issue: the rapid dissemination of fake news. Fake news, prevalent on social media and various online platforms, poses a significant threat as it can cause substantial social and national harm. It can be used as a tool for propaganda against individuals, societies, organizations, or political parties. Additionally, spammers exploit enticing news headlines to generate revenue through advertisements, employing tactics like click-baits. To address this problem, this study focuses on binary classification of online news articles using Artificial Intelligence, Natural Language Processing, and Machine Learning techniques. By analyzing existing research in fake news detection, we aim to identify the most effective traditional machine learning models. Our goal is to develop a product model using supervised machine learning algorithms that can accurately classify news articles as true or false. We utilize tools such as Python’s scikit-learn and NLP for textual analysis. The objective is to empower users with the ability to distinguish between fake and genuine news, while also verifying the authenticity of the websites publishing the news.

Keywords- Machine Learning, Click Baits, NLP, Online News.

1. INTRODUCTION

   In the digital age, the pervasive spread of fake news through online platforms and social media has emerged as a major societal concern. With the internet revolutionizing news dissemination, the distinction between accurate reporting and misinformation has become increasingly blurred. This research endeavors to tackle this issue by harnessing advanced technologies such as Artificial Intelligence (AI), Natural Language Processing (NLP), and Machine Learning. These tools are pivotal in analyzing extensive textual data, enabling the identification of linguistic nuances and aiding in the differentiation between authentic news articles and fabricated content. The study meticulously explores and refines traditional machine learning models, incorporating data mining techniques to extract pertinent features from news articles. The primary objective of this research is to empower individuals to critically evaluate online information. By providing practical solutions rooted in AI and machine learning, this study promotes media literacy and informed decision-making. Through the development of robust algorithms, this research aims to equip users with the

   ability to discern the authenticity of news sources, fostering a more discerning and knowledgeable society. Ultimately, this interdisciplinary approach contributes valuable insights to the field of fake news detection, offering practical tools to address the challenges posed by misinformation in the digital era.

   In addition to the technological advancements, this study delves into the social and ethical implications of fake news. By understanding the motivations behind the creation and dissemination of false information, the research provides a comprehensive perspective on combating this digital menace. Moreover, the research investigates the integration of website authenticity checks, ensuring a holistic approach to verifying the credibility of sources. This multifaceted strategy aims not only to enhance media literacy but also to fortify the digital landscape against the detrimental effects of misinformation, fostering a more trustworthy and informed online environment.

2. LITERATURE REVIEW [1] In this paper, (Aditya Darak et al. 2018) employs the J48 Algorithm, a machine learning decision tree method, to detect fake product reviews. Data mining techniques are utilized to extract hidden patterns within reviews, enabling the system to identify suspicious behaviors. Java programming language, NetBeans IDE, and MySQL database management system are integrated to create a robust and efficient fraud detection system. This interdisciplinary approach showcases the synergy of machine learning, data mining, and database technologies in combating fake product reviews in e- commerce. The study emphasizes the significance of advanced algorithms and real-time data analysis to ensure the accuracy and reliability of review classification.
   [2] This paper, (Shuo Yu et al. 2019) presents REAL, a novel approach for detecting fake reviewer groups in online review systems. REAL employs modularity-based graph clustering with Graph Convolutional Networks (GCNs) to find candidate groups. It combines group-level and individual- level anomaly indicators to measure the suspiciousness of each group. The proposed method effectively balances group size and precision, outperforming three baseline methods in detecting suspicious reviewer groups. The study emphasizes the importance of identifying collusive fake reviewer groups, which can significantly impact the reputation of products in online review platforms.

   Published by : http://www.ijert.org

   [3]The paper (Lu Yuan et al. 2023) presents a comprehensive overview of fake news detection methodologies, categorizing fake news into different types and discussing detection methods from various angles. The research employs interdisciplinary approaches, incorporating insights from psychology, neuro-cognitive science, linguistics, and communication science. It explores content-based detection methods, social network-based methods, and knowledge- based methods. Additionally, the study delves into multimodal fake news detection, combining text, images, and other data types for enhanced accuracy. The research emphasizes the integration of knowledge and data in the detection process, involving natural language processing, computer vision, and deep learning techniques.
   [4] The paper, (Lu Huang et al. 2023) proposes an approach for fake news detection using deep learning techniques. It introduces FNFNet, a deep convolutional neural network model, designed to automatically learn features for accurate fake news detection. FNFNet outperforms existing methods, achieving a high accuracy rate. The model utilizes parallel convolutional layers, max-pooling, dense layers, and ReLU activation functions to process textual data efficiently. This approach significantly improves fake news detection by leveraging the power of deep learning algorithms.
   [5]In the paper titled “AENeT: an Attention-Enabled Neural Architecture for Fake News Detection using Contextual Features,” the authors address the growing challenge of fake news in the era of social media. They propose a deep learning model that combines contextual embeddings, attention mechanisms, and relevant metadata to assess the degree of fakeness in news statements. The model integrates ELMo embeddings and GloVe embeddings, leveraging attention mechanisms to weigh important words in the sentence. The proposed architecture achieved an accuracy of 46.36% on the LIAR dataset, surpassing the current state-of-the-art by 1.49%. This research presents an effective solution for detecting fake news by leveraging deep learning techniques and contextual features.
   [6] In this paper, (Linmei Hu et al. 2023) proposes CLIMB, a novel framework for multi-modal fake news detection, considering the image-text matching bias through causal inference. It formulates the fake news detection task as a causal graph, addressing the confounding effect of image-text matching during training. CLIMB uses a vision-language pre- trained model (ViLT) to estimate the image-text matching degree and a two-tower co-attention model for basic fake news detection. The framework leverages backdoor adjustment to remove the confounding bias, enabling accurate classification. Extensive experiments on real-world datasets validate the effectiveness of CLIMB in improving multi- modal fake news detection.

   International Journal of Engineering Research & Technology (IJERT)

   ISSN: 2278-0181

   Volume 13, Issue 01 January 2024

   [7] In this paper, (Boutheina Jlif et al. 2022) the authors propose a Soft Three-Level Voting Model (Soft T-LVM) for COVID-19 fake news detection. The model consists of three levels: data preprocessing and feature extraction, soft voting at multiple levels, and hyperparameter tuning. The authors use a dataset obtained from a competition and employ techniques like lowercasing, punctuation removal, tokenization, stopword removal, stemming, and lemmatization for data preprocessing. Three feature extraction techniques, including Bag-of-words, TF-IDF, and N-grams, are used to vectorize the input data. The authors experiment with five machine learning models: SVM, RF, LR, KNN, and Multinomial NB, using different combinations of feature extraction vectors. Soft voting is applied at each level to combine the outputs of individual models, and the process stops after three levels. The proposed Soft T-LVM achieves an accuracy of 94.43%, precision of 94.56%, recall of 94.43%, and F1-score of 94.43%, outperforming other models and existing works in COVID-19 fake news detection. The AUC score of Soft T-LVM is also very high, indicating the model’s excellent predictive performance.
   [8] The research paper titled “Incorporating User- Comment Graph for Fake News Detection” proposes a novel framework called Graph Comment-User Advanced Learning (GCAL) for detecting fake news. The method combines textual content analysis, user-comment interactions, and graph neural networks. It constructs a heterogeneous graph representing users, comments, and their relationships. Utilizing attention mechanisms, the model captures semantic affinity between news sentences and user comments, enhancing fake news detection accuracy. The proposed GCAL outperforms existing methods, showcasing its effectiveness in identifying fake news in social media contexts.
   [9] In the study conducted by Jain et al., the authors introduce an advanced neural architecture, AENeT, for discerning the authenticity of news statements in the era of rampant fake news. Their model integrates contextual embeddings, attention mechanisms, and pertinent metadata to enhance the accuracy of fake news detection. Utilizing pre-trained word embeddings such as GloVe and contextualized embeddings like ELMo, the model captures intricate semantic meanings within words and sentences. Through attention mechanisms, the model emphasizes crucial words and metadata, elevating its performance. Remarkably, the proposed architecture achieves a significant accuracy of 46.36% on the LIAR dataset, surpassing existing methods by 1.49%. This research presents a noteworthy contribution to the realm of fake news detection, harnessing the power of deep learning techniques.
   [10] The study by Dharmaraj R. Patil addresses the challenge of identifying fake news amidst the abundance of online information. Using a Kaggle dataset containing 20,800 news articles, the authors propose a Majority Voting approach, combining machine learning classifiers like Decision Tree, Logistic Regression, and others. Through advanced pre-processing techniques and feature extraction using TF-IDF and CountVectorizer, the method achieves an impressive accuracy of 96.38%, outperforming individual classifiers.

   The Majority Voting technique proves effective in distinguishing between real and fake news, providing a robust solution for combating misinformation online.

3. EXISTING SYSTEM

   In the realm of fake news detection, several methodologies and technologies have been harnessed to combat the spread of misinformation. Natural Language Processing (NLP) techniques play a pivotal role, enabling the analysis of textual data to identify linguistic patterns, sentiment, and contextual cues that may indicate the presence of fake news. NLP algorithms, such as text classification and sentiment analysis, are employed to assess the authenticity of news articles. Additionally, network analysis techniques are utilized to study the dissemination patterns of fake news on social media platforms. By examining the flow of information through social networks, researchers can identify suspicious patterns of sharing and propagation, aiding in the detection of misinformation campaigns. Fact-checking databases and external knowledge bases are integrated into fake news detection systems. These resources provide a reference point for validating the claims made in news articles. Automated fact-checking algorithms compare the information in news stories against these databases, flagging content that contradicts established facts. User behavior analysis is another facet of existing fake news detection systems. By studying how users interact with news content online, analysts can discern patterns in user behavior, such as excessive sharing or engagement with dubious sources. Unusual behavioral patterns can raise red flags, indicating potential instances of fake news. Moreover, efforts have been made to incorporate multimedia analysis into fake news detection. With the rise of manipulated images and videos, algorithms capable of detecting visual manipulations are deployed. Image forensics and video analysis techniques help identify edited or fabricated visual content, enhancing the overall accuracy of fake news detection systems. Despite the progress, challenges persist. The rapid evolution of fake news tactics demands continuous advancements in detection methods. Adversarial attacks, where misinformation is deliberately crafted to evade detection, pose a significant challenge. Researchers are exploring machine learning models robust to such attacks, ensuring the resilience of fake news detection systems. The existing systems for fake news detection employ a multifaceted approach, integrating NLP, network analysis, fact-checking databases, user behavior analysis, and multimedia forensics. While these methods have proven effective, ongoing research and innovation are essential to stay ahead of the evolving landscape of fake news. While significant strides have been made in fake news detection, several challenges persist, underscoring the need for continuous research and innovation. One of the primary challenges is the sheer volume of data circulating online. The vast amount of information makes it difficult to identify and verify the authenticity of every piece of news in real-time. Scalable solutions, powered by advanced algorithms and parallel computing, are essential to handle the ever-increasing volume of online content. Additionally, the global and diverse nature of fake news presents a challenge. Misinformation can

   vary widely across cultures, languages, and contexts, making it crucial to develop detection models that are sensitive to these nuances. Cross-lingual and cross-cultural analysis, coupled with context-aware algorithms, are pivotal in ensuring the effectiveness of fake news detection systems across diverse societal landscapes. The rapid evolution of technology also poses a challenge. As new mediums of communication emerge, such as deepfake videos and AI- generated text, fake news detection methods must adapt. Researchers are exploring innovative techniques, including deep learning algorithms and generative adversarial networks (GANs), to detect sophisticated and visually convincing fake content. Furthermore, user privacy and ethical considerations are paramount in the development of fake news detection systems. Striking a balance between ensuring user privacy and collecting the necessary data for analysis is a complex ethical dilemma. Researchers are actively exploring privacy- preserving techniques, such as federated learning, where models are trained across decentralized devices without exchanging raw data, thus safeguarding user privacy while advancing detection capabilities. In the future, interdisciplinary collaboration will play a vital role in enhancing fake news detection. Collaboration between computer scientists, social scientists, linguists, and ethicists can provide holistic insights into the multifaceted challenges of fake news. By fostering interdisciplinary partnerships, researchers can develop comprehensive solutions that address not only the technical aspects but also the social and ethical dimensions of misinformation. In conclusion, the fight against fake news is an ongoing battle that requires continuous innovation, interdisciplinary collaboration, and a commitment to ethical practices. As technology advances, so too must our strategies for detecting and combating misinformation, ensuring the integrity of information in the digital age.

4. PROPOSED SYSTEM

In the initial stage of Data Collection, once the dataset is collected, it is preprocessed using various sentiment analysis techniques. When working on fake news detection, extracting linguistic features from the text can provide valuable insights for training the model. Additionally, performing feature selection helps identify the most informative and relevant features for the task. A novel approach called PICO has been used wherein a variety of machine learning algorithms that include the supervised, unsupervised, reinforcement machine learning algorithms are used for Classification purposes.

They have to be trained with a data set called train data set. The algorithms used by fake news detection systems include machine learning algorithms such as Logistic Regression, Random Forests, Decision trees, and Gradient Boosting.

Architecture diagram:

Fig 1.1

IV MODULES

Module 1: Data gathering and preprocessing

In this process, datasets are collected from kaggle.com. It involves supervised learning as the dataset contains both input and output parameters. Perform data cleaning to remove noise and ensure data quality. This includes:

• Removing irrelevant information or columns that are not

  useful for your task.

• Handling missing data by either imputing values or removing instances with missing data, depending on the impact on the dataset’s integrity.
• Addressing inconsistencies, such as inconsistent formatting

  or encoding issues in the text.

  Preprocess the textual content of the news articles to prepare them for analysis. This typically involves the following steps:

• Load Raw Data: Read and load the raw text data that you’ve

  collected from various sources.

• Text Cleaning: Clean the text data to remove any unwanted elements that might not contribute to fake news detection. This includes removing HTML tags, special characters, punctuation, URLs, and numbers.
• Tokenization: Split the text into individual words or tokens. Tokenization helps break down the text into smaller units, making it easier for further processing.
• Stopword Removal: Remove common words (stopwords) that don’t carry significant meaning, such as “and,” “the,” “is,” etc. These words can be safely discarded to reduce noise in the data.
• Lemmatization or Stemming: Reduce words to their base or root form. Lemmatization considers the meaning of the word, while stemming simply chops off prefixes or suffixes. This step helps to consolidate variations of the same word.
• Text Vectorization: Convert the processed text into numerical form that can be fed into machine learning algorithms.
• Feature Scaling: If your chosen algorithm requires it, scale the numerical features to have a similar scale, which can improve convergence and performance.

  Module 2: Classification using Logistic Regression

  Logistic Regression is a popular algorithm for binary classification problems. It models the relationship between the input features and the probability of belonging to a particular class. Logistic Regression uses a logistic function (also called the sigmoid function) to map the input features to a probability score between 0 and 1. It estimates the coefficients for each feature using a method like maximum likelihood estimation. The decision boundary is determined by a threshold value, typically 0.5, above which the instance is classified as one class and below which it is classified as the other class.

• Preprocess the textual features of the news articles, such as

  using bag-of-words

  or TF-IDF representations.

• Split the dataset into training and testing sets.
• Train the logistic regression model on the training set, adjusting the model parameters to minimize the logistic loss or maximize the log-likelihood.
• Evaluate the model’s performance on the testing set using

  metrics like accuracy, precision, recall, or F1 score.

  Module 3: Decision Tree Classification

  Decision Trees are non-parametric supervised learning algorithms that use a hierarchical structure of decisions and conditions to classify instances. Decision Trees recursively partition the feature space based on the values of input features, creating a tree-like structure. At each internal node, a decision is made based on a specific feature and its value, leading to different branches. Leaf nodes represent the final class assignments based on the features’ values and the decision boundaries defined by the tree structure.

• Preprocess the textual features of the news articles, similar

  to logistic regression.

• Split the dataset into training and testing sets.
• Train the decision tree model on the training set, using an

  algorithm like ID3.

• Tune hyperparameters, such as the maximum tree depth or minimum number of samples per leaf node, to prevent overfitting.
• Evaluate the model’s performance on the testing set using

  classification metrics.

  Module 4: Random Forest Classifier

  In this module, we have used a powerful and widely used ensemble method Random Forest for classification. It combines the predictions of multiple decision trees to make accurate and robust predictions. Random Forest is an ensemble learning method that constructs a collection of decision trees and combines their predictions through voting or averaging. Random Forest creates an ensemble of decision trees by randomly selecting subsets of the original dataset (bootstrapping) and features at each tree’s

  construction. Each decision tree is trained on a different subset of the data, introducing randomness and reducing the risk of overfitting. During prediction, each decision tree in the forest independently classifies the instance, and the final

  prediction is determined by majority voting (for classification problems) or averaging (for regression problems) over

  all the trees’ predictions.

  Module 5: Gradient Boosting Classifier

  Gradient Boosting is a boosting algorithm that sequentially builds an ensemble of weak prediction models, typically decision trees, to create a strong predictive model.

  The algorithm starts by fitting an initial model to the training data. In each subsequent iteration, the algorithm focuses on the instances that were not well predicted by the

  previous models, placing more emphasis on them. It fits a new model to these instances, adjusting the model’s parameters to minimize the loss function with respect to the previous predictions. The final model is an ensemble of weak models, where each model’s contribution is determined by a learning rate and the model’s performance on the training data.

• Preprocess the textual features of the news articles, such as

  using bag-of-words or TF-IDF representations.

• Split the dataset into training and testing sets.
• Train a Gradient Boosting classifier (e.g., GBM or XGBoost) on the training set, specifying the number of boosting iterations and other hyperparameters.
• Gradient Boosting classifiers typically provide feature importance scores, indicating the relative importance of each feature in the classification process.
• Evaluate the model’s performance on the testing set using

  classification metrics.

  Module 6: Model Testing

  Model testing is a crucial step in evaluating the performance and effectiveness of a fake news detection model. It involves assessing how well the trained model generalizes to unseen data. Here’s an outline of the process for testing a fake news detection model:

  Test Set Preparation and Model Prediction:

  This includes preparing a separate dataset, referred to as the test set, that was not used during training or model development. This ensures that the test set represents

  unseen data. And then preprocessing the test set using the same steps applied to the training set, including text preprocessing and feature extraction. The trained fake news detection model is applied to the test set. The predictions for each instance in the test set is obtained, which indicates whether it is classified as real or fake news.

  Evaluation Metrics:

  The various evaluation metrics are calculated to assess the model’s performance. Common evaluation metrics for binary classification tasks include:

• Accuracy: The proportion of correctly classified instances.
• Precision: The proportion of true positives out of all predicted positives. It measures the model’s ability to correctly identify fake news.
• Recall: The proportion of true positives out of all actual positives. It measures the model’s ability to capture all instances of fake news.
• F1 Score: The harmonic mean of precision and recall,

  providing a balanced measure of performance.

  Module 7: Flask Framework

  Flask is a Python web framework for creating dynamic web applications.It enables a user-friendly interface for your fake news detection model.Seamlessly integrates machine learning into a web app without users needing technical knowledge.Offers simplicity, modularity, and smooth communication between user browsers and the app.

  Flask Web App Components:

• HTML Templates: Define page layout and dynamic content

  presentation.

• Routes: Map URLs to Python functions, handling user

  interactions.

• Static Files: Provide visual elements like CSS, images, and JavaScript. These components create a cohesive, interactive, and engaging user experience.

User Interaction:

User inputs news text in the provided textbox. Web app sends input to fake news detection model. Model predicts news authenticity (real or fake). Result displayed on the web page. Seamless process simplifies user interaction and provides quick insights.

V RESULTS

VI CONCLUSION

Fake news can take many forms, including text, images, and videos, making it challenging to develop models that can detect various types of misinformation. Fake news exists in multiple languages, and detecting it in languages with limited resources can be difficult. It has the ability to spread rapidly on social media platforms, making it difficult to detect and counteract before it reaches a wide audience. Development of tools and plugins that allow users to verify information before sharing it can be beneficial. Machine learning and natural language processing techniques are improving, offering better tools for fake news detection. Educating the public about fake news and critical thinking reduces the impact of bogus information.

REFERENCES:

[1] A. Dabhade, S. Roy, M. S. Moustafa, S. A. Mohamed, R. El Gendy and S. Barma, “Extreme Weather Event (Cyclone) Detection in India Using Advanced Deep Learning Techniques,” 2021 9th International Conference on Orange Technology (ICOT), Tainan, Taiwan, 2021, pp. 1-4, doi: 10.1109/ICOT54518.2021.9680663.

[2] A. Nair et al., “A Deep Learning Framework for the Detection of Tropical Cyclones From Satellite Images,” in IEEE Geoscience and Remote Sensing Letters, vol. 19, pp.

1-5, 2022, Art no. 1004405, doi: 10.1109/LGRS.2021.3131638.

[3] J. S. Combinido, J. R. Mendoza and J. Aborot, “A Convolutional Neural Network Approach for Estimating Tropical Cyclone Intensity Using Satellite-based Infrared Images,” 2018 24th International Conference on Pattern Recognition (ICPR), Beijing, China, 2018, pp. 1474-1480, doi: 10.1109/ICPR.2018.8545593.

[4] X. Yu, Z. Chen, G. Chen, H. Zhang and J. Zhou, “A Tensor Network for Tropical Cyclone Wind Speed Estimation,” IGARSS 2019 – 2019 IEEE International Geoscience and Remote Sensing Symposium, Yokohama, Japan, 2019, pp. 10007-10010, doi: 10.1109/IGARSS.2019.8898937.

[5] W. Tian, X. Zhou, W. Huang, Y. Zhang, P. Zhang and S. Hao, “Tropical Cyclone Intensity Estimation Using Multidimensional Convolutional Neural Network From Multichannel Satellite Imagery,” in IEEE Geoscience and Remote Sensing Letters, vol. 19, pp. 1-5, 2022, Art no. 5511105, doi: 10.1109/LGRS.2021.3134007.

[6] S. Shakya, S. Kumar and M. Goswami, “Deep Learning Algorithm for Satellite Imaging Based Cyclone Detection,” in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 13, pp. 827-839, 2020, doi: 10.1109/ JSTARS.2020.2970253.

[7] P. Wang, P. Wang, D. Wang and B. Xue, “A Conformal Regressor With Random Forests for Tropical Cyclone Intensity Estimation,” in IEEE

Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1-14, 2022, Art no. 4105914, doi: 10.1109/TGRS.2021.3139930.

[8] R. Zhang, Q. Liu and R. Hang, “Tropical Cyclone Intensity Estimation Using Two-Branch Convolutional Neural Network From Infrared and Water Vapor Images,” in IEEE Transactions on Geoscience and Remote Sensing, vol. 58, no. 1, pp. 586-597, Jan. 2020, doi: 10.1109/TGRS.2019.2938204.

[9] F. Meng, Q. Tian, H. Sun, D. Xu and T. Song, “Cyclone Identify using Two- Branch Convolutional Neural Network from Global Forecasting System Analysis,” 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS, Brussels, Belgium, 2021, pp. 8468-8471, doi: 10.1109/IGARSS47720.2021.9554662.

[10] R. Carmo, N. Longépé, A. Mouche, D. Amorosi and N. Cremer, “Deep Learning Approach for Tropical Cyclones Classification Based on C-Band Sentinel-1 SAR Images,” 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS, 2021.

[11] Aditya Darak , Akhil Chaudhary , Prajwal Mogaveera, Fake Product Review Detection and Removal, in International Journal of Science and Research (IJSR) ISSN: 2319-7064 Volume 7 Issue 10, October 2018.

[12] Chen Cao, Shihao Li, Shuo Yu, Zhikui Chen, Fake Reviewer Group Detection in Online Review Systems arXiv:2112.06403v1 [cs.SI] 13 Dec 2021.

[13] Lu Yuan, Hangshun Jiang, Hao Shen, Lei Shi, and Nanchang Cheng Sustainable Development of Information Dissemination: A Review of Current Fake News Detection Research and Practice Systems 2023, 11, 458. https://doi.org/10.3390/systems11090458.

[14] Vidit Jain, Rohit Kaliyar, Anurag Goswami, Pratik Narang, AENeT: an attention-enabled neural architecture for fake news detection using contextual features in Neural Computing and Applications (Vol. 34, ssue 1, pp. 771782). Springer Science and Business Media LLC.

[15] Lu Huang Deep Learning for Fake News Detection: Theories and Models

EITCE 2022, October 2123, 2022.

[16] Linmei Hu, Ziwei Chen, Ziwang Zhao, Jianhua Yin, Liqiang Nie, Causal Inference for Leveraging Image-Text Matching Bias in Multi-Modal Fake News Detection in: IEEE Transactions on Knowledge and Data Engineering ( Volume: 35, Issue: 11, 01 November 2023)

[17] Boutheina Jlifi, Chayma Sakrani & Claude Duvallet, Towards a soft three- level voting model (Soft T-LVM) for fake news detection in Journal of Intelligent Information Systems 2023.

[18] Hao Liao, Qixin Liu, Kai Shu, Xing Xie, Incorporating User-Comment Graph for Fake News Detection in arXiv:2011.01579v1 [cs.SI] 3 Nov 2020.

[19] Vidit Jain, Rohit Kumar Kaliyar, Anurag Goswami, Pratik Narang,and Yashvardhan Sharma, AENeT: an attention-enabled neural architecture for fake news detection using contextual features published online 2021 Aug.

[20] Dharmaraj R. Patil, Fake News Detection Using Majority Voting Technique published online Mar 2022.