Abstract The rise of fake news has become a critical issue for information integrity, particularly in low-resource languages, which often lack sufficient digital data and tools for machine learning (ML) and deep learning (DL) models. This systematic review examines the effectiveness of ML and DL techniques in detecting fake news across multiple languages, with particular emphasis on the research gap concerning low-resource languages, in contrast to the extensive studies focused on monolingual English. We conducted a comprehensive search across several databases, screening 1,567 records and including 85 studies in our final analysis, based on well-defined inclusion and exclusion criteria. Additionally, the review explores various definitions of fake news and rumors, publicly available datasets, and commonly employed evaluation tools in detection methods. We provide a thorough analysis of both traditional and advanced ML and DL techniques, highlighting key challenges and potential avenues for future research. While these advanced models have led to significant improvements, they are not without limitations. For instance, transformer models, despite their power, may inadvertently capture biases from training data, potentially affecting their performance across different domains or languages. Hybrid models, while enhancing capabilities, may face challenges related to computational costs and scalability. Furthermore, the dependence on large datasets and complex architectures can limit the practicality of these models for fake news detection (FND) in low-resource settings or real-time applications. Consequently, while the integration of advanced models and features has advanced FND, ongoing research is needed to address these challenges and improve model applicability in diverse contexts. Future work should focus on mitigating biases, improving model efficiency, and developing methods to adapt these models for lower-resource environments and real-time scenarios. This study provides a comprehensive roadmap for future research aimed at overcoming these challenges and advancing FND across diverse linguistic contexts.

Abstract Early detection of skin cancer from skin lesion images using visual inspection can be challenging. In recent years, research in applying deep learning models to assist in the diagnosis of skin cancer has achieved impressive results. State-of-the-art techniques have shown high accuracy, sensitivity and specificity compared with dermatologists. However, the analysis of dermoscopy images with deep learning models still faces several challenges, including image segmentation, noise filtering and image capture environment inconsistency. After making the introduction to the topic, this paper firstly presents the components of machine learning-based skin cancer diagnosis. It then presents the literature review on the current advance in machine learning approaches for skin cancer classification, which covers both the traditional machine learning approaches and deep learning approaches. The paper also presents the current challenges and future directions for skin cancer classification using machine learning approaches.

Abstract Fault detection in belt conveyor idlers is crucial for minimising downtime and reducing maintenance costs in industrial operations. Traditional methods, like vibration or temperature-based monitoring, face limitations, including challenging sensor installation and restricted data accessibility. Moreover, these approaches often emphasise spatial features, neglecting the temporal dynamics essential for understanding idler performance over time. This study introduces TD-CLNet, a hybrid fault detection framework that leverages acoustic signals captured via contactless microphones processed through a Time-Distributed CNN-LSTM architecture. The model combines the spatial feature extraction capabilities of Convolutional Neural Networks (CNNs) with the temporal sequence modelling strengths of Long Short-Term Memory (LSTM) networks. A key innovation is the use of the Time-Distributed layer, which enables consistent feature extraction across individual log-Mel spectrogram frames while preserving their temporal relationships. This ensures a robust and coordinated learning process, efficiently addressing the challenges of detecting complementary and relevant features. The performance of TD-CLNet is compared to a frame-based feature extraction approach, which treats each log-Mel spectrogram frame as an independent sample, as well as traditional machine learning methods. Results demonstrate that TD-CLNet achieves a test accuracy of 92% on real-world idler data using K-fold cross-validation, significantly outperforming competing methods. This research provides a scalable and effective solution for fault detection in belt conveyor idlers, advancing predictive maintenance strategies, improving operational efficiency, and minimising unplanned downtime in industrial environments.

The rise of social media as a source of news consumption has led to the spread of fake news, posing serious consequences for both individuals and society. The detection and prevention of fake news are essential, and previous research has shown that incorporating news content along with its associated headlines and user comments can improve detection performance. However, the semantic relationships between these elements have not been fully explored. This paper proposes a novel approach that models the relationships between news bodies and associated headlines/user comments using deep learning techniques , such as fine-tuned Bidirectional Encoder Representations from Transformers (BERT) and cross-level cross-modality attention sub-networks. In our proposed model, we utilize two different configurations of BERT: pool-based representation, which provides a representation of the entire document, and sequence representation, which represents each token within the document (i.e., at the word and text levels). The approach also encodes user-posting behavioural features and fuses the output of these components to detect fake news using a classification layer. Our experiments on benchmark datasets demonstrate the superiority of the proposed method over existing state-of-the-art (SOTA) approaches, highlighting the importance of utilizing semantic relationships for improved fake news detection (FND). These findings have significant implications for combating the spread of fake news and protecting society from its negative effects.

Conventional malware detection approaches have the overhead of feature extraction, the requirement of domain experts, and are time-consuming and resource-intensive. Learning-based approaches are the mainstay of malware detection as they overcome most of these challenges by significantly improving the detection effectiveness and providing a low false positive rate. The exponential growth of malware variants and first-time-appeared malware, which includes polymorphic and zero-day attacks, are some of the significant challenges to learning-based malware detectors. These challenges have catastrophic impacts on the detection effectiveness of these learning-based malware detectors. This paper proposes a novel deep learning-based framework to detect first-time-appeared malware effectively and efficiently by providing better performance than conventional malware detection approaches. First, it translates and visualises each Windows portable executable (PE) file into a coloured image to eliminate the overhead of feature extraction and the need for domain experts to analyse the features. In the subsequent step, a fine-tuned deep learning model is used to extract the deep features from the last fully connected layer. The step has reduced the cost of training required by the deep learning models if used for end-to-end classification. The third step selects the most important and influential features through a powerful feature selection algorithm. The most important features are then fed to a one-class classifier for final detection. With the one-class classifier, an enclosed boundary around the features of benign data is constructed. Anything outside the boundary is declared as an anomaly/malicious. It has enhanced the framework's ability to detect evolving, unseen, polymorphic, and zero-day attacks, as well as reducing the problem of overfitting. The detection effectiveness of the proposed framework is validated with state-of-the-art deep learning models and conventional approaches. The proposed framework has outperformed with an accuracy of 99.30% on the Malimg dataset. The Wilcoxon signed-rank test is used to validate the statistical significance of the proposed framework. It is evident from the results that the proposed framework is effective and can be used in the defence industry, resulting in more powerful and robust solutions against zero-day and polymorphic attacks.

The rise of social media platforms has led to a proliferation of false information in various forms. Identifying malicious entities on these platforms is challenging due to the complexities of natural language and the sheer volume of textual data. Compounding this difficulty is the ability of these entities to deliberately modify their writing style to make false information appear trustworthy. In this study, we propose a neural-based framework that leverages the hierarchical structure of input text to detect both fake news content and fake news spreaders. Our approach utilizes enhanced Hierarchical Convolutional Attention Networks (eHCAN), which incorporates both style-based and sentiment-based features to enhance model performance. Our results show that eHCAN outperforms several strong baseline methods, highlighting the effectiveness of integrating deep learning (DL) with stylistic features. Additionally, the framework uses attention weights to identify the most critical words and sentences, providing a clear explanation for the model’s predictions. eHCAN not only demonstrates exceptional performance but also offers robust evidence to support its predictions.

The proliferation of fake news across languages and domains on social media platforms poses a significant societal threat. Current automatic detection methods for low-resource languages (e.g., Swahili, Indonesian and other low-resource languages) face limitations due to two factors: sequential length restrictions in pre-trained language models (PLMs) like multilingual bidirectional encoder representation from transformers (mBERT), and the presence of noisy training data. This work proposes a novel and efficient multilingual fake news detection (MFND) approach that addresses these challenges. Our solution leverages a hybrid extractive and abstractive summarization strategy to extract only the most relevant content from news articles. This significantly reduces data length while preserving crucial information for fake news classification. The pre-processed data is then fed into mBERT for classification. Extensive evaluations on a publicly available multilingual dataset demonstrate the superiority of our approach compared to state-of-the-art (SOTA) methods. Our analysis, both quantitative and qualitative, highlights the strengths of this method, achieving new performance benchmarks and emphasizing the impact of content condensation on model accuracy and efficiency. This framework paves the way for faster, more accurate MFND, fostering more robust information ecosystems.

Abstract The proliferation of fake news on social media platforms poses significant challenges to society and individuals, leading to negative impacts. As the tactics employed by purveyors of fake news continue to evolve, there is an urgent need for automatic fake news detection (FND) to mitigate its adverse social consequences. Machine learning (ML) and deep learning (DL) techniques have emerged as promising approaches for characterising and identifying fake news content. This paper presents an extensive review of previous studies aiming to understand and combat the dissemination of fake news. The review begins by exploring the definitions of fake news proposed in the literature and delves into related terms and psychological and scientific theories that shed light on why people believe and disseminate fake news. Subsequently, advanced ML and DL techniques for FND are dicussed in detail, focusing on three main feature categories: content-based, context-based, and hybrid-based features. Additionally, the review summarises the characteristics of fake news, commonly used datasets, and the methodologies employed in existing studies. Furthermore, the review identifies the challenges current FND studies encounter and highlights areas that require further investigation in future research. By offering a comprehensive overview of the field, this survey aims to serve as a guide for researchers working on FND, providing valuable insights for developing effective FND mechanisms in the era of technological advancements.

Abstract Disease diagnosis is an exciting task due to many associated factors. Inaccuracy in the measurement of a patient’s symptoms and the medical expert’s expertise has some limitations capacity to articulate cause affects the diagnosis process when several connected variables contribute to uncertainty in the diagnosis process. In this case, a decision support system that can assist clinicians in developing a more accurate diagnosis has a lot of potentials. This work aims to deploy a fuzzy inference-based decision support system to diagnose various diseases. Our suggested method distinguishes new cases based on illness symptoms. Distinguishing symptomatic disorders becomes a time-consuming task in most cases. It is critical to design a system that can accurately track symptoms to identify diseases using a fuzzy inference system (FIS). Different coefficients were used to predict and compute the severity of the predicted diseases for each sign of disease. This study aims to differentiate and diagnose COVID-19, typhoid, malaria and pneumonia. The FIS approach was utilized in this study to determine the condition correlating with input symptoms. The FIS method demonstrates that afflictive illness can be diagnosed based on the symptoms. Our decision support system’s findings showed that FIS might be used to identify a variety of ailments. Doctors, patients, medical practitioners and other healthcare professionals could benefit from our suggested decision support system for better diagnosis and treatment.

The COVID-19 pandemic has resulted in a surge of fake news, creating public health risks. However, developing an effective way to detect such news is challenging, especially when published news involves mixing true and false information. Detecting COVID-19 fake news has become a critical task in the field of natural language processing (NLP). This paper explores the effectiveness of several machine learning algorithms and fine-tuning pre-trained transformer-based models, including Bidirectional Encoder Representations from Transformers (BERT) and COVID-Twitter-BERT (CT-BERT), for COVID-19 fake news detection. We evaluate the performance of different downstream neural network structures, such as CNN and BiGRU layers, added on top of BERT and CT-BERT with frozen or unfrozen parameters. Our experiments on a real-world COVID-19 fake news dataset demonstrate that incorporating BiGRU on top of the CT-BERT model achieves outstanding performance, with a state-of-the-art F1 score of 98%. These results have significant implications for mitigating the spread of COVID-19 misinformation and highlight the potential of advanced machine learning models for fake news detection.

Abstract Skin cancer can be fatal if it is found to be malignant. Modern diagnosis of skin cancer heavily relies on visual inspection through clinical screening, dermoscopy, or histopathological examinations. However, due to similarity among cancer types, it is usually challenging to identify the type of skin cancer, especially at its early stages. Deep learning techniques have been developed over the last few years and have achieved success in helping to improve the accuracy of diagnosis and classification. However, the latest deep learning algorithms still do not provide ideal classification accuracy. To further improve the performance of classification accuracy, this paper presents a novel method of classifying skin cancer in clinical skin images. The method consists of four blocks. First, class rebalancing is applied to the images of seven skin cancer types for better classification performance. Second, an image is preprocessed by being split into patches of the same size and then flattened into a series of tokens. Third, a transformer encoder is used to process the flattened patches. The transformer encoder consists of N identical layers with each layer containing two sublayers. Sublayer one is a multihead self-attention unit, and sublayer two is a fully connected feed-forward network unit. For each of the two sublayers, a normalization operation is applied to its input, and a residual connection of its input and its output is calculated. Finally, a classification block is implemented after the transformer encoder. The block consists of a flattened layer and a dense layer with batch normalization. Transfer learning is implemented to build the whole network, where the ImageNet dataset is used to pretrain the network and the HAM10000 dataset is used to fine-tune the network. Experiments have shown that the method has achieved a classification accuracy of 94.1%, outperforming the current state-of-the-art model IRv2 with soft attention on the same training and testing datasets. On the Edinburgh DERMOFIT dataset also, the method has better performance compared with baseline models.

The dissemination of fake news on the Internet has resulted in worrying negative implications for individuals and society. This paper begins by discussing the definitions of fake news and the related terms that have often co-occurred with the term fake news. Then, we summarised several social science theories characterising fake news spreading. Next, we discussed the state-of-the-art techniques for detecting fake news using news content and user context information. Finally, we conducted a case study that demonstrates that the interplay between news content and context-based features helps uncover useful patterns to discriminate fake from real news. Our study suggests that content and context-based features are necessary for better performance of fake news detection.

Abstract In today’s world, lung cancer is a significant health burden, and it is one of the most leading causes of death. A leading type of lung cancer is malignant mesothelioma (MM). Most of the MM patients do not show any symptoms. Etiology plays a vital factor in the diagnosis of any disease. Positron emission tomography (PET), magnetic resonance imaging (MRI), biopsies, X-rays and blood tests are essential but costly and invasive MM risk factor identification methods. In this work, we mainly focused on the exploration of the MM risk factors. The identification of mesothelioma symptoms was carried out by utilizing the data of mesothelioma patients. However, the dataset was comprised of both healthy and mesothelioma patients. The dataset is prone to a class imbalance problem in which the number of MM patients significantly less than healthy individuals. To overcome the class imbalance problem, the synthetic minority oversampling technique has been utilized. The association rule mining-based Apriori algorithm has been applied to a preprocessed dataset. Before using the Apriori algorithm, both duplicate and irrelevant attributes were removed. Moreover, the numerical attributes were also classified into nominal attributes and the association rules were generated in the dataset. Our results show that erythrocyte sedimentation rate, asbestos exposure and its duration time, and pleural and serum lactic dehydrogenase ratio are major risk factors of MM. The severe stages of MM can be avoided by earlier identification of risk factors of the disease. The failure of identification of risk factors can lead to increased risk of multiple medical conditions, including cardiovascular diseases, mental distress, diabetes and anemia.

Abstract The area of corporate bankruptcy prediction attains high economic importance, as it affects many stakeholders. The prediction of corporate bankruptcy has been extensively studied in economics, accounting and decision sciences over the past two decades. The corporate bankruptcy prediction has been a matter of talk among academic literature and professional researchers throughout the world. Different traditional approaches were suggested based on hypothesis testing and statistical modeling. Therefore, the primary purpose of the research is to come up with a model that can estimate the probability of corporate bankruptcy by evaluating its occurrence of failure using different machine learning models. As the dataset was not well prepared and contains missing values, various data mining and data pre-processing techniques were utilized for data preparation. Within this research, the task of resolving the issues induced by the imbalance between the two classes is approached by applying different data balancing techniques. We address the problem of imbalanced data with the random undersampling and Synthetic Minority Over Sampling Technique (SMOTE). We used five machine learning models (support vector machine, J48 decision tree, Logistic model tree, random forest and decision forest) to predict corporate bankruptcy earlier to the occurrence. We use data from 2009 to 2013 on Poland manufacturing corporates and selected the 64 financial indicators to be broken down. The main finding of the study is a significant improvement in predictive accuracy using machine learning techniques. We also include other economic indicators ratios, along with Altman’s Z-score variables related to profitability, liquidity, leverage and solvency (short/long term) to propose an efficient model. Machine learning models give better results while balancing the data through SMOTE as compared to random undersampling. The machine learning technique related to decision forest led to 99% accuracy, whereas support vector machine (SVM), J48 decision tree, Logistic Model Tree (LMT) and Random Forest (RF) led to 92%, 92.3%, 93.8% and 98.7% accuracy, respectively, with all predictive financial indicators. We find that the decision forest outperforms the other techniques and previous techniques discussed in the literature. The proposed method is also deployed on the web to assist regulators, investors, creditors and scholars to predict corporate bankruptcy.

Lung cancer is one of the most deadly diseases. It has a high death rate and its incidence rate has been increasing all over the world. Lung cancer appears as a solitary nodule in chest x-ray radiograph (CXR). Therefore, lung nodule detection in CXR could have a significant impact on early detection of lung cancer. Radiologists define a lung nodule in CXR as "solitary white nodule-like blob." However, the solitary feature has not been employed for lung nodule detection before. In this paper, a solitary feature-based lung nodule detection method was proposed. We employed stationary wavelet transform and convergence index filter to extract the texture features and used AdaBoost to generate white nodule-likeness map. A solitary feature was defined to evaluate the isolation degree of candidates. Both the isolation degree and the white nodule likeness were used as final evaluation of lung nodule candidates. The proposed method shows better performance and robustness than those reported in previous research. More than 80% and 93% of lung nodules in the lung field in the Japanese Society of Radiological Technology (JSRT) database were detected when the false positives per image were two and five, respectively. The proposed approach has the potential of being used in clinical practice.

AdaBoost classifiers with Haar-like features are widely used for license plate (LP) localization. However, it normally requires high-dimensional Haar-like features which cause extremely high computational cost. In this paper, a rejection cascade was built for LP localization with reduced Haar-like features. We first introduced line segment features as pre-input of Haar-like features for AdaBoost to eliminate more than 70% of the background in an image. Line segment features, including density, directionality and regularity, were extracted from line segments, which were detected by applying Hough Transform on an edge image. Later, AdaBoost classifiers with Haar-like features were further applied to identify the exact location of license plates. Our method dramatically reduced the demanded dimensions of Haar-like features, therefore saved much time in AdaBoost training stage. By comparing our method with methods of only using Haar-like features and only using line segment features, experimental results demonstrated that our proposed method achieved the best detection rate with significantly reduced dimensions of Haar-like features.

Packaging appearance is extremely important in cigarette manufacturing. Typically, there are two types of cigarette packaging defects: (1) cigarette laying defects such as incorrect cigarette numbers and irregular layout; (2) tin paper handle defects such as folded paper handles. In this paper, an automated vision-based defect inspection system is designed for cigarettes packaged in tin containers. The first type of defects is inspected by counting the number of cigarettes in a tin container. First k-means clustering is performed to segment cigarette regions. After noise filtering, valid cigarette regions are identified by estimating individual cigarette area using linear regression. The k clustering centers and area estimation function are learned off-line on training images. The second kind of defect is detected by checking the segmented paper handle region. Experimental results on 500 test images demonstrate the effectiveness of the proposed inspection system. The proposed method also contributes to the general detection and classification system such as identifying mitosis in early diagnosis of cervical cancer.

Emotional factors directly reflect audiences' attention, evaluation and memory. Affective contents analysis not only create an index for users to access their interested movie segments, but also provide feasible entry for video highlights. Most of the work focus on emotion type detection. Besides emotion type, emotion intensity is also a significant clue for users to find their interested content. For some film genres (Horror, Action, etc), the segments with high emotion intensity have the most possibilities to be video highlights. In this paper, we propose a hierarchical structure for emotion categories and analyze emotion intensity and emotion type by using arousal and valence related features hierarchically. Firstly, High, Medium and Low are detected as emotion intensity levels by using fuzzy c-mean clustering on arousal features. Fuzzy clustering provides a mathematical model to represent vagueness, which is close to human perception. After that, valence related features are used to detect emotion types (Anger, Sad, Fear, Happy and Neutral). Considering video is continuous time series data and the occurrence of a certain emotion is affected by recent emotional history, Hidden Markov Models (HMMs) are used to capture the context information. Experimental results shows the movie segments with high emotion intensity cover over 80% of the movie highlights in Horror and Action movies and the hierarchical method outperforms the one-step method on emotion type detection. Meanwhile, it is flexible for user to pick up their favorite affective content by choosing both emotion intensity levels and emotion types.

Due to the wide viewer-ship and high commercial potentials, recently, sports video analysis attracts extensive research efforts. One of the main tasks in sports video analysis is to identify sports genres i.e. sports video categorization. Most of the existing work focus on mapping content-based features to sports genres by using supervised learning methods. Moreover, video data sets seeks efficient data reduction methods due to the large size and noisy data. It lacks comparison analysis on the implementation and performance of these methods. In this paper, the research is carried out by using four dominant machine learning algorithms, namely Decision Tree, Support Vector Machine, K Nearest Neighbor and Naive Bayesian, and comparing their performance on a high dimensional feature set which selected by some feature selection tools such as Correlation-based Feature Selection (CFS), Principal Components Analysis (PCA) and Relief. Experimental results shows that Support Vector Machine (SVM) and k-NN are not sensitive to reduction of training sets. Moreover, three different feature reduction methods perform very differently with respect to four different tools.

Sports video has attracted a global viewership. Research effort in this area has been focused on semantic event detection in sports video to facilitate accessing and browsing. Most of the event detection methods in sports video are based on visual features. However, being a significant component of sports video, audio may also play an important role in semantic event detection. In this paper, we have borrowed the concept of the “keyword” from the text mining domain to define a set of specific audio sounds. These specific audio sounds refer to a set of game-specific sounds with strong relationships to the actions of players, referees, commentators, and audience, which are the reference points for interesting sports events. Unlike low-level features, audio keywords can be considered as a mid-level representation, able to facilitate high-level analysis from the semantic concept point of view. Audio keywords are created from low-level audio features with learning by support vector machines. With the help of video shots, the created audio keywords can be used to detect semantic events in sports video by Hidden Markov Model (HMM) learning. Experiments on creating audio keywords and, subsequently, event detection based on audio keywords have been very encouraging. Based on the experimental results, we believe that the audio keyword is an effective representation that is able to achieve satisfying results for event detection in sports video. Application in three sports types demonstrates the practicality of the proposed method.

The paper describes a novel approach for classifying continuous speech into visible mouth-shape related classes (called visemes). The selection and comparison of various acoustic speech features and the use of context information in the classification are addressed. Continuous speech is classified into 9 visible mouth-shape related classes on an acoustic frame basis. Some mouth-shape related acoustic speech signal features are selected as the input to a classifier constructed with recurrent neural network (RNN). 304 training sentences and 88 testing sentences are chosen from DARPA TIMIT continuous speech database. The average viseme recognition rate for the test set reaches 84.7% on frame level, which is a quite promising result considering that the test is applied on continuous multi-speakers and large vocabulary speech.>