Figure - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
There have been many efforts to detect rumors using various machine learning (ML) models, but there is still a lack of understanding of their performance against different rumor topics and available features, resulting in a significant performance degrade against completely new and unseen (unknown) rumors. To address this issue, we investigate the...
Contexts in source publication
Context 1
... better understand each of the rumor events, we show some key basic statistics in Table 5. In general, the non-rumor (NR) tweets have a higher favorite, retweet and friends counts (average 1.84, 1.09 and 1.12 times higher, respectively, without "Ebola & Gurlitt" event), as well as the number of keywords used and the number of unique keywords (labeled Unique in the table) across all tweets. ...Context 2
... and user-based features were not sufficiently effective in rumor detection in our experiments. As shown in Table 5, rumor events had quite different base metrics that contribute towards the propagation and user-based features. However, the rumor detection performance by ML models was not affected by them. ...Similar publications
Health complications during the gestation period have evolved as a global issue. These complications sometimes result in the mortality of the fetus, which is more prevalent in developing and underdeveloped countries. The genesis of Machine Learning (ML) algorithms in the healthcare domain have brought remarkable progress in disease diagnosis, treat...
![Confusion matrix for multi-class classification [21]](publication/363099952/figure/tbl1/AS:11431281081753700@1661871639889/Confusion-matrix-for-multi-class-classification-21_Q320.jpg)
Purpose: However, from the variety of uses of these algorithms, in general, accuracy problems are still a concern today, even accuracy problems related to multi-class classification still require further research.Methods: This study proposes a stacking ensemble classifier method to produce better accuracy by combining Logistic Regression, Random Fo...
Depression detection from social media texts such as Tweets or Facebook comments could be very beneficial as early detection of depression may even avoid extreme consequences of long-term depression i.e. suicide. In this study, depression intensity classification is performed using a labeled Twitter dataset. Further, this study makes a detailed per...
Issues of providing mental health support to people with emerging or current mental health disorders are becoming a significant concern throughout the world. One of the biggest effects of digital psychiatry during COVID-19 is its capacity for early identification and forecasting of a person's mental health decline resulting in chronic mental health...
Machine learning techniques play an increasingly prominent role in medical diagnosis. With the use of these techniques, patients’ data can be analyzed to find patterns or facts that are difficult to explain, making diagnoses more reliable and convenient. The purpose of this research was to compare the efficiency of diabetic classification models us...
Citations
... Por se tratarem de modelos de aprendizagem profunda, em geral o classificador em sié composto por algumas camadas do tipo feed-forward seguidas por uma função softmax, com os esforços direcionados a produzir características latentes independentes de domínio. Entre as exceções temos [Kim et al. 2020] com o uso de ensembles de algoritmos clássicos de aprendizado de máquina e [Kong et al. 2023] que utiliza de programação genética para produzir uma equação matemática para detecção de fake news. ...
... A variação dentro de eventos distintos, como Covid-19 e Monkeypox, tambémé considerada, expondo a dificuldade de encontrar um modelo abrangente o suficiente para funcionar com todos os tipos de fake news, inclusive as emergentes . Além disso, vários estudos analisados concentraram-se na detecção de fake news relacionadasà pandemia da Covid-19 [Zhu et al. 2023, Omrani et al. 2023, Kong et al. 2023, Kim et al. 2020, Rastogi et al. 2021, evidenciando a necessidade de combater notícias falsas emergentes em tempos de crise de saúde pública. ...
... Para incrementar a precisão na detecção de fake news, algumas estratégias incluíram a incorporação de informações adicionais, tais como a credibilidade das fontes de notícias [Birunda and Devi 2021], análises de conteúdo histórico e interações de usuários [Tang et al. 2023], bem como o aproveitamento de bases de conhecimento [Zhang et al. 2019, Lu et al. 2022. Ainda, características como sentimentos, polaridade e atributos manualmente definidos continuam a ser relevantes [Zhu et al. 2023, Kong et al. 2023, Ras-togi et al. 2021, Sicilia et al. 2021, Gautam and Jerripothula 2020, Kim et al. 2020]. ...
Este artigo apresenta uma revisão sistemática sobre a detecção de fake news em domínios cruzados, em que o desafioé identificar desinformação em contextos variados, como diferentes temas, idiomas ou fontes. A revisão revela uma preferência pela Generalização de Domínio (DG), que busca desen-volver modelos capazes de identificar fake news em uma ampla gama de contex-tos sem ajustes específicos, em detrimento da Adaptação de Domínio (DA), que visa otimizar o desempenho de um modelo treinado em um domínio fonte para domínios-alvo específicos. A diversidade dos conjuntos de dados utilizados res-salta a necessidade de benchmarks padronizados para avaliações consistentes. O estudo sugere a exploração de novas técnicas de generalização e adaptação de domínio para aprimorar a detecção de fake news em diferentes contextos.
... However, using all of these features may not improve the performance of the news classifier. The utilization of one or more of these features is based on the type of fake news detection issue that must be addressed [28]. A few studies indicated that experiments showed that network and user features were not sufficient for detecting rumors. ...
... For news validation, news content (linguistics and visual data) is used as a feature in fake news detection models [30]. The results of research conducted by Kim, Kim [28] demonstrated that the accuracy of detecting rumors using only the content-based feature was higher than using all other features simultaneously combined. Human psychology favors articles with appealing multimedia materials coupled with text since readers believe them. ...
Currently, social networks have become the main source to acquire news about current global affairs. However, fake news appears and spreads on social media daily. This disinformation has a negative influence on several domains, such as politics, the economy, and health. In addition, it further generates detriments to societal stability. Several studies have provided effective models for detecting fake news in social networks through a variety of methods; however, there are limitations. Furthermore, since it is a critical field, the accuracy of the detection models was found to be notably insufficient. Although many review articles have addressed the repercussions of fake news, most have focused on specific and recurring aspects of fake news detection models. For example, the majority of reviews have primarily focused on dividing datasets, features, and classifiers used in this field by type. The limitations of the datasets, their features, how these features are fused, and the impact of all these factors on detection models were not investigated, especially since most detection models were based on a supervised learning approach. This review article analyzes relevant studies for the few last years and highlights the challenges faced by fake news detection models and their impact on their performance. The investigation of fake news detection studies relied on the following aspects and their impact on detection accuracy, namely datasets, overfitting/underfitting, image-based features, feature vector representation, machine learning models, and data fusion. Based on the analysis of relevant studies, the review showed that these issues significantly affect the performance and accuracy of detection models. This review aims to provide room for other researchers in the future to improve fake news detection models.
... However, the proliferation of fake news is on the rise, which calls for further development and exploration of new directions in research to improve the techniques used for identifying such news [5,18]. Many studies on the identification of fake news on social networks depend on one or more features such as content, network propagation, or user [19][20][21]. Analyzing users' comments to ascertain their attitudes toward the news could play a major role in identifying fake news [22][23][24] and giving an idea of the credibility of the published news [14,15]. Albahar [25] posited that user comments have great discriminatory value in the detection of fake news, wherein the expression of sentiment [26] or emotion [27] is crucial. ...
... Sometimes features used in other studies may not provide a significant improvement in the performance of the detection models, or they may contain noise that reduces their performance. Sometimes, the auxiliary features used in previous studies may not provide a significant improvement in the performance of the model, as a study Kim, Kim [21] indicated that the accuracy of identifying rumors using content features only was higher than using all features combined at the same time. In the future, we plan to analyze the emotions embedded in news and find out if fake news contains certain emotions that represent the publisher's stance, in addition to analyzing the emotions of user comments by exploring other types of emotions that represent the public's responses to such news. ...
Nowadays, social media has become the main source of news around the world. The spread of fake news on social networks has become a serious global issue, damaging many aspects, such as political, economic, and social aspects, and negatively affecting the lives of citizens. Fake news often carries negative sentiments, and the public’s response to it carries the emotions of surprise, fear, and disgust. In this article, we extracted features based on sentiment analysis of news articles and emotion analysis of users’ comments regarding this news. These features were fed, along with the content feature of the news, to the proposed bidirectional long short-term memory model to detect fake news. We used the standard Fakeddit dataset that contains news titles and comments posted regarding them to train and test the proposed model. The suggested model, using extracted features, provided a high detection accuracy of 96.77% of the Area under the ROC Curve measure, which is higher than what other state-of-the-art studies offer. The results prove that the features extracted based on sentiment analysis of news, which represents the publisher’s stance, and emotion analysis of comments, which represent the crowd’s stance, contribute to raising the efficiency of the detection model.
... Depending on the nature of the issue, one or more features may be used. The results of the study by Kim, Kim [54] indicated that rumor detection accuracy using user features was the most poor of all. In contrast, rumor detection accuracy using content-only features was significantly higher than utilizing all features at once. ...
... Although some of the results from some of the datasets examined may be encouraging, the majority of the results are poor. Kim, Kim [54] proposed an Ensemble Solution (ES) based on Soft Voting which consists of RF, XGBoost, and Multilayer perception. In order to determine which ML model was most suitable for creating an ES model, they examined several. ...
In recent times, social media has become the primary way people get news about what is happening in the world. Fake news surfaces on social media every day. Fake news on social media has harmed several domains, including politics, the economy, and health. Additionally, it has negatively affected society's stability. There are still certain limitations and challenges even though numerous studies have offered useful models for identifying fake news in social networks using many techniques. Moreover, the accuracy of detection models is still notably poor given we deal with a critical topic. Despite many review articles, most previously concentrated on certain and repeated sections of fake news detection models. For instance, the majority of reviews in this discipline only mentioned datasets or categorized them according to labels, content, and domain. Since the majority of detection models are built using a supervised learning method, it has not been investigated how the limitations of these datasets affect detection models. This review article highlights the most significant components of the fake news detection model and the main challenges it faces. Data augmentation, feature extraction, and data fusion are some of the approaches explored in this review to improve detection accuracy. Moreover, it discusses the most prominent techniques used in detection models and their main advantages and disadvantages. This review aims to help other researchers improve fake news detection models.
... In [3], the researchers conducted experiments on ensemble models and created a model comprising random forest (RF), extreme gradient boosting (XGBoost), and a multilayer perceptron. The performance achieved by this ensemble model outperformed that of classical machine learning models by 20%. ...
With the increased popularity of social media platforms, people are increasingly depending on them for news and updates. Even official media channels post news on social media platforms such as Twitter and Facebook. However, with the vast amount of user-generated content, the credibility of shared information must be verified, and this process should be performed automatically and efficiently to accommodate the huge rate of generated posts. Current technology provides powerful methods and tools to solve the issue of rumor spreading on social networks. In this study, the aim is to investigate the use of state-of-the-art machine learning and deep learning models to detect rumors in a collection of Arabic tweets using the ArCOV19-Rumors dataset. A comprehensive comparison of the performance of the models was conducted. In deep learning experiments, the performances of seven optimizers were compared. The results demonstrated that using over-sampled data did not enhance classical and deep learning models. By contrast, using stacking classifiers increased the predictive model’s performance. As a result, the model became more logical and realistic in predicting rumors, non-rumors, and other classes than using classical machine learning without the stacking technique. Additionally, both long short-term memory (LSTM) and bidirectional-LSTM (Bi-LSTM) with the Root mean square propagation (RMSprop) optimizer obtained the best results. Finally, the results were analyzed to explain and interpret the low performance.
Currently, the Internet ranks first among sources of information. In the recent period, the role of online social networks (OSN) has significantly increased, which has both positive and negative consequences. The negative role of OSN is related to the spread of fake news that affects people's daily lives, manipulates their thoughts and feelings, changes their beliefs and can lead to wrong decisions. The problem of spreading fake news in OSN is currently global, and the formation of countermeasures is an urgent task today. Today, there are various proven approaches to detecting fake news. In particular, one of the approaches is based on the use of different machine (ML) and deep (DL) learning algorithms. The other is based on the results of sentiment analysis of news content and analysis of emotions in user comments. The research conducted by the authors of other approaches to detecting fake news, which differ from the ones given, made it possible to conclude that the mentioned approaches are effective and promising in terms of using their potential for the development of new models with high performance indicators on various data sets. In the article, the author's ideas regarding the improvement of existing approaches to detecting fake news based on the use of the potential of these approaches are formed and formalized. The first idea is based on the implementation of the mechanism of combining machine (ML) and deep (DL) learning methods, as well as the results of the analysis of the sentiment of news content and emotions in user comments, which takes into account the possibility of ensuring a sufficient level of effectiveness in detecting fake news, a certain level of the values of the selected metrics, as well as a certain level of functional characteristics of the author's method. The second idea is based on the implementation of a mechanism combining the functionality of two methods from among the specified two groups, which would provide optimal parameters for detecting fake news according to defined criteria and indicators. The substantiation of the ideas involved the preliminary implementation of: setting the researched problem; functional analysis of machine (ML) and deep (DL) learning algorithms, as well as fake news detection algorithms based on the use of the results of sentiment analysis of news content and emotions in user comments; description of metrics for evaluating the effectiveness of methods for detecting fake news. According to the results of the substantiation of the perspective of the ideas, the tasks of detecting fake news in the author's production were formalized.
With the emergence of online social networks, the dissemination and acquisition of information have experienced dramatic transformations. While social media makes peoples’ lives easier, it also speeds up the creation and spread of rumors. Therefore, solving the problem of reliably and effectively recognizing words has become a critical need. The global–local attention network (GLAN)-based rumor detection model has been improved to improve its accuracy. Considering the impact of the positioning relationship among words in text on rumor identification, a new relative positional encoding method is utilized to enhance the original model’s local feature extraction module. This method can more precisely extract the semantic and location information of the text in a rumor and aggregate it to provide a better text feature that differentiates rumors from nonrumors. This characteristic is combined with the global element specifying forwarding behavior to increase the effectiveness of word detection. Experimental findings demonstrate that the
F
1
value of the suggested method on the Weibo dataset may reach 95%, with a unique detection effect compared to other mainstream detection methods.
The struggle of organisations from all fields to find practical answers for identifying online-based fake news is a prevalent problem right now. The news is published on news Websites, which act as official sources. Social media has drawn the attention of individuals from all over the world who use it to disseminate fake news because of its accessibility, cost, and ease of information exchange. Because people are unable to distinguish between true and misleading information, fake news weakens the logic of the truth, endangering democracy, journalism, and public confidence in political institutions. To sustain strong Internet media and informal organisations, fake news detection must be automated. The manual method is now impractical, slow, expensive, very subjective, and biased due to the vast quantity of data that is available on social networks. As a result, an interesting and fruitful area of research is automated data categorisation. ML and DL algorithms are by far the best way for fake news detection. This study provided an exhaustive, insightful, and empirical assessment encompassing all AI strategies for the recognising fake news, including reinforcement learning, ensemble learning, unsupervised learning, supervised learning, and semi-supervised learning.KeywordsDeep learning(DL)Fake news classificationMachine learning (ML)Social media
