Search

Machine learning techniques that rely on textual features or sentiment lexicons can lead to erroneous sentiment analysis. These techniques are especially vulnerable to domain-related difficulties, especially when dealing in Big data. In addition, labeling is time-consuming and supervised machine learning algorithms often lack labeled data. Transfer learning can help save time and obtain high performance with fewer datasets in this field. To cope this, we used a transfer learning-based Multi-Domain Sentiment Classification (MDSC) technique. We are able to identify the sentiment polarity of text in a target domain that is unlabeled by looking at reviews in a labelled source domain. This research aims to evaluate the impact of domain adaptation and measure the extent to which transfer learning enhances sentiment analysis outcomes. We employed transfer learning models BERT, RoBERTa, ELECTRA, and ULMFiT to improve the performance in sentiment analysis. We analyzed sentiment through various transformer models and compared the performance of LSTM and CNN. The experiments are carried on five publicly available sentiment analysis datasets, namely Hotel Reviews (HR), Movie Reviews (MR), Sentiment140 Tweets (ST), Citation Sentiment Corpus (CSC), and Bioinformatics Citation Corpus (BCC), to adapt multi-target domains. The performance of numerous models employing transfer learning from diverse datasets demonstrating how various factors influence the outputs. [ABSTRACT FROM AUTHOR]

The current limitations of user–product interaction with smart speakers have spurred the proposal of a model to circumvent these challenges. We used the ViP design principle to redefine the user's approach to interacting with the product. Throughout the deconstruction and design stages, we explored the structure and function of the conventional product across three layers: the product layer, interaction layer, and context layer using three models. We used the hierarchical analysis method to effectively quantify the design factors affecting user experience and identify the key design factors. This approach enabled us to contextualize the smart audio system, explore the interaction dynamics between the product and the user, and provide valuable insights on designing new products. A questionnaire method was used to survey 67 users, and a reliability test was conducted to ensure the validity of the questionnaire v (Cronbach's coefficient α = 0.868). A pairwise comparison of factors was conducted on a 1–9 scale, with weights determined through the analytic hierarchy process (AHP). The combination of the ViP design principle and hierarchical analysis presents a novel and objective paradigm to guide designers to customize product characteristics (design attributes) to enhance user human–computer interaction experience. We validated the feasibility of the innovative design approach using the smart speaker model, offering insights for research on designing similar products. [ABSTRACT FROM AUTHOR]

Image tempering is one of the significant issues in the modern era. The use of powerful tools for image editing with advanced technology and its widespread on social media raised questions on data integrity. Currently, the protection of images is uncertain and a severe concern, mainly when it transfers over the Internet. Thus, it is essential to detect an anomaly in images through artificial intelligence techniques. The simple way of image forgery is called copy-move, where a part of an image is replicated in the same image to hide unwanted content of the image. However, image processing through handcrafted features usually looks for pattern concerns with duplicate content, limiting their employment for huge data classification. On the other side, deep learning approaches achieve promising results, but their performance depends on training data with fine-tuning of hyperparameters. Thus, we proposed a custom convolutional neural network (CNN) architecture with a pre-trained model ResNet101 through a transfer learning approach. For this purpose, both models are trained on five different datasets. In both cases, the impact of the model is evaluated through accuracy, precision, recall, F -score and achieved the highest 98.4% accuracy using the Coverage dataset. [ABSTRACT FROM AUTHOR]

Cybersecurity threats are increasing rapidly as hackers use advanced techniques. As a result, cybersecurity has now a significant factor in protecting organizational limits. Intrusion detection systems (IDSs) are used in networks to flag serious issues during network management, including identifying malicious traffic, which is a challenge. It remains an open contest over how to learn features in IDS since current approaches use deep learning methods. Hybrid learning, which combines swarm intelligence and evolution, is gaining attention for further improvement against cyber threats. In this study, we employed a PSO-GA (fusion of particle swarm optimization (PSO) and genetic algorithm (GA)) for feature selection on the CICIDS-2017 dataset. To achieve better accuracy, we proposed a hybrid model called LSTM-GRU of deep learning that fused the GRU (gated recurrent unit) and LSTM (long short-term memory). The results show considerable improvement, detecting several network attackswith 98.86% accuracy. A comparative study with other current methods confirms the efficacy of our proposed IDS scheme. [ABSTRACT FROM AUTHOR]

Viruses infect millions of people worldwide each year, and some can lead to cancer or increase the risk of cancer. As viruses have highly mutable genomes, new viruses may emerge in the future, such as COVID-19 and influenza. Traditional virology relies on predefined rules to identify viruses, but new viruses may be completely or partially divergent from the reference genome, rendering statistical methods and similarity calculations insufficient for all genome sequences. Identifying DNA/RNA-based viral sequences is a crucial step in differentiating different types of lethal pathogens, including their variants and strains. While various tools in bioinformatics can align them, expert biologists are required to interpret the results. Computational virology is a scientific field that studies viruses, their origins, and drug discovery, where machine learning plays a crucial role in extracting domain- and task-specific features to tackle this challenge. This paper proposes a genome analysis system that uses advanced deep learning to identify dozens of viruses. The system uses nucleotide sequences from the NCBI GenBank database and a BERT tokenizer to extract features from the sequences by breaking them down into tokens. We also generated synthetic data for viruses with small sample sizes. The proposed system has two components: a scratch BERT architecture specifically designed for DNA analysis, which is used to learn the next codons unsupervised, and a classifier that identifies important features and understands the relationship between genotype and phenotype. Our system achieved an accuracy of 97.69% in identifying viral sequences. [ABSTRACT FROM AUTHOR]

Modern technologies such as the Internet of Things (IoT) and physical systems used as navigation systems play an important role in locating a specific location in an unfamiliar environment. Due to recent technological developments, users can now incorporate these systems into mobile devices, which has a positive impact on the acceptance of navigational systems and the number of users who use them. The system that is used to find a specific location within a building is known as an indoor navigation system. In this study, we present a novel approach to adaptable and changeable multistory navigation systems that can be implemented in different environments such as libraries, grocery stores, shopping malls, and official buildings using facial and speech recognition with the help of voice broadcasting. We chose a library building for the experiment to help registered users find a specific book on different building floors. In the proposed system, to help the users, robots are placed on each floor of the building, communicating with each other, and with the person who needs navigational help. The proposed system uses an Android platform that consists of two separate applications: one for administration to add or remove settings and data, which in turn builds an environment map, while the second application is deployed on robots that interact with the users. The developed system was tested using two methods, namely system evaluation, and user evaluation. The evaluation of the system is based on the results of voice and face recognition by the user, and the model's performance relies on accuracy values obtained by testing out various values for the neural network parameters. The evaluation method adopted by the proposed system achieved an accuracy of 97.92% and 97.88% for both of the tasks. The user evaluation method using the developed Android applications was tested on multi-story libraries, and the results were obtained by gathering responses from users who interacted with the applications for navigation, such as to find a specific book. Almost all the users find it useful to have robots placed on each floor of the building for giving specific directions with automatic recognition and recall of what a person is searching for. The evaluation results show that the proposed system can be implemented in different environments, which shows its effectiveness. [ABSTRACT FROM AUTHOR]

Female accounts for approximately 50% of the total population worldwide and many of them had breast cancer. Computer‐aided diagnosis frameworks could reduce the number of needless biopsies and the workload of radiologists. This research aims to detect benign and malignant tumors automatically using breast ultrasound (BUS) images. Accordingly, two pretrained deep convolutional neural network (CNN) models were employed for transfer learning using BUS images like AlexNet and DenseNet201. A total of 697 BUS images containing benign and malignant tumors are preprocessed and performed classification tasks using the transfer learning‐based CNN models. The classification accuracy of the benign and malignant tasks is completed and achieved 92.8% accuracy using the DensNet201 model. The results thus achieved compared in state of the art using benchmark data set and concluded proposed model outperforms in accuracy from first stage breast tumor diagnosis. Finally, the proposed model could help radiologists diagnose benign and malignant tumors swiftly by screening suspected patients. [ABSTRACT FROM AUTHOR]