IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
Artificial intelligence-driven method for the discovery and prevention of distributed denial of service attacks
Get PDFDistributed denial of service (DDoS) attacks has emerged as a prominent cyber threat in contemporary times. By impeding the machine's capacity to give services to legitimate clients, the impacted system performance and buffer size are reduced. Researchers are working to build sophisticated algorithms that can identify and thwart DDoS violations. An effective approach for DDoS attacks has been proposed in this work. This research presents a model as a potential explanation for DDoS assaults. In order to successfully identify this kind of attacks, which may stop or block the urgent and vital transmission of data, we present a distinctive method that integrates a pair of fully connected layers within an amalgamated deep learning (DL) framework with long short-term memory (LSTM) and a max pooling layer. The acquired accuracy reached 99.58%
User sentiment dynamics in social media: a comparative analysis of X and Threads
Get PDFThis research examines the dynamics of user sentiment and its correlation with the usage factors of applications in the context of the competition between X (formerly Twitter) and Threads, a social media application under the umbrella of Meta. Through sentiment analysis of user reviews on the Google Play Store and App Store, the study aims to identify the key factors contributing to a significant decline in user engagement with Threads and the return of users to X. The method employed in this research is the support vector machine (SVM) for sentiment classification of reviews. The study then correlates the classified sentiments with application usage factors: usability, features, design, and support. The research findings indicate user sentiment influences user engagement, especially in features and design. The research concludes with insights regarding implications for application developers and suggests directions for future research
Large language models-based metric for generative question answering systems
Get PDFIn the evolving landscape of text generation, which has advanced rapidly in recent years, techniques for evaluating the performance and quality of the generated text lag behind relatively. Traditionally, lexical-based metrics such as bilingual evaluation understudy (BLEU), recall-oriented understudy for gisting evaluation (ROUGE), metric for evaluation of translation with explicit ordering (METEOR), consensus-based image description evaluation (CIDER), and F1 have been utilized, primarily relying on n-gram similarity for evaluation. In recent years, neural and machine-learning-based metrics, like bidirectional encoder representations from transformers (BERT) score, key phrase question answering (KPQA), and BERT supervised training of learned evaluation metric for reading comprehension (LERC) have shown superior performance over traditional metrics but suffered from a lack of generalization towards different domains and requires massive human-labeled training data. The main contribution of the current research is to investigate the use of train-free large language models (LLMs) as scoring metrics, evaluators, and judges within a questionanswering context, encompassing both closed and open-QA scenarios. To validate this idea, we employ a simple zero-shot prompting of Mixtral 8x7 B, a popular and widely used open-source LLM, to score a variety of datasets and domains. The experimental results on ten different benchmark datasets are compared against human judgments, revealing that, on average, simple LLMbased metrics outperformed sophisticated state-of-the-art statistical and neural machine-learning-based metrics by 2-8 points on answer-pairs scoring tasks and up to 15 points on contrastive preferential tasks
Predicting water resistance and pitching angle during take-off: an artificial neural network approach
Get PDFThis research addresses the challenges faced by seaplanes and amphibious aircraft during takeoff and landing on water, emphasizing the limitations and costs associated with traditional towing tank tests and computational fluid dynamics (CFD) simulations. The study proposes an innovative approach that employs artificial neural networks (ANN) to predict water resistance and pitching angle during amphibious aircraft take-off, minimizing the reliance on expensive towing tank tests. The ANN models are developed and optimized using Bayesian optimization, showcasing improved accuracy in predicting water resistance and pitching angle. The research demonstrates the potential of machine learning, specifically ANNs, to significantly reduce the need for costly experimental tests, providing an efficient alternative for designing amphibious aircraft. The results indicate high accuracy in predicting water resistance and pitching angle, offering substantial time and resource savings during the experimental phase. However, the study highlights the need for model adaptation for different designs and test variations to enhance overall applicability
Recommender system for dengue prevention using machine learning
Get PDFThe study aimed to develop a recommender system for dengue prevention using environmental factors and mosquito larvae data. Data were collected from 100 households in Surat Thani, Thailand using mosquito larval survey in January 2020. Data mining techniques: frequent pattern growth (FP-Growth) and Apriori algorithms were used to find association rules and to compare accuracies for selecting a suitable model. The recommender system was designed as a web application. FP-Growth is more suitable for these data than Apriori algorithm. The factors associated with dengue infection, including community area, densely populated area, and agricultural area. Most areas where mosquito larvae are found are community areas and agricultural areas. Aedes larvae were found most in water containers with dark colors and without a lid. Aedes larvae were also found in small water jars, large water jars, cement tanks, and plastic tanks. The recommender system should be useful to dengue vector prevention and to health service communities, in planning and operational activities
Uncertainty-aware contextual multi-armed bandits for recommendations in e-commerce
Get PDFThe growing e-commerce landscape has seen a shift towards personalized product recommendations, which play a critical role in influencing consumer behavior and driving revenue. This study explores the efficacy of contextual multi-armed bandits (CMAB) in optimizing personalized recommendations by intelligently balancing exploration and exploitation. Recognizing the inherent uncertainty in user behaviors, we propose an enhanced CMAB policy that incorporates item correlation matrix as an additional component of uncertainty to the conventional binary exploration and exploitation setup of bandit policies. Our approach aims to increase the overall relevance of recommendations through the 'triadic framework’ of CMAB, that seamlessly integrates with existing bandit policies, enabling adaptive recommendations based on diverse user attributes. By outperforming traditional models, this uncertainty-aware method demonstrates its potential in refining recommendation accuracy, thus maximizing revenue in a competitive e-commerce environment. Future research will explore dynamic uncertainty modeling and cross-domain applications to further advance the field
Empowering SDN with DDoS attack detection: leveraging hybrid machine learning based IDPS controller for robust security
Get PDFSoftware-defined network (SDN) is an innovative networking framework where a centralized controller manages networking administration and sorts out network traffic issues. It becomes difficult for the controller to identify the malicious user who is sending a large number of spoofed packets, such as in a distributed denial of service (DDoS) attack. To prevent DDoS attacks from damaging legitimate users, it is important to take steps to prevent them. The issue of preventing DDoS attacks in SDN remains unresolved despite many algorithms proposed. Methods presented in this paper employ bandwidth threshold estimation, which triggers the intrusion detection and prevention system (IDPS) controller if the threshold is exceeded. Whenever the threshold is exceeded due to network congestion, transferred packets are filtered at the server level by identifying the utilization of bandwidth in OpenDaylight (ODL) and POX. K-nearest neighbor (K-NN) and support vector machine (SVM) are used by the IDPS controller to detect and thwart DDoS attacks. Using Mininet, two SDN centralized controllers are simulated to improve performance significantly. Based on SVM in the ODL controller, this work has provided mitigation techniques for preventing DDoS attacks with an accuracy of 96.75% compared to previously published accuracy
A novel light-weight convolutional neural network for rice leaf disease classification
Get PDFRice is one of the primary sources of staple meals. It may turn out to be a disaster as the production of agricultural products is declined due to diseases and therefore it is required to straighten up the situation by taking precautionary measures. Generally, deep learning (DL) architectures are employed for the identification of plant leaf diseases and it is observed that there is a trade-off between the accuracy and parameters. This study introduces a light-weight architecture called rice leaf disease classification convolutional neural network (RLDC-CNN). The objective of the proposed architecture is to improve the accuracy and reduce the loss by using a combination of convolutional layers, maxpooling layers, and fully connected layers. These layers use activation function for non-linearity, dropout for regularization and implements hyperparameter tuning with various optimizers that include Adam, RMSprop, stochastic gradient descent (SGD) and adaptive gradient (AdaGrad). Experiments are conducted on the dataset of 7,096 images with batch size of 32 under various learning rates. The behavior is analyzed by comparing the existing models and the count of parameters (in millions) equipped by RLDC-CNN, DenseNet121, VGG-16, and ResNet50 is 0.65, 8.49, 15.44, 26.49 with accuracy of 99.15%, 98.94%, 97.82%, 96.48% respectively
Convolutional neural network based encoder-decoder for efficient real-time object detection
Get PDFConvolutional neural networks (CNN) are applied to a variety of computer vision problems, such as object recognition, image classification, semantic segmentation, and many others. One of the most important and difficult issues in computer vision, object detection, has attracted a lot of attention lately. Object detection validating the occurrence of the object in the picture or video and then properly locating it for recognition. However, under certain circumstances, such as when an item has issues like occlusion, distortion, or small size, there may still be subpar detection performance. This work aims to propose an efficient deep learning model with CNN and encoder decoder for efficient object detection. The proposed model is experimented on Microsoft Common Objects in Context (MS-COCO) dataset and achieved mean average precision (mAP) of about 54.1% and accuracy of 99%. The investigational outcomes amply showed that the suggested mechanism could achieve a high detection efficiency compared with the existing techniques and needed little computational resources
Robust two-stage object detection using YOLOv5 for enhancing tomato leaf disease detection
Get PDFDeep learning facilitates human activities across various sectors, including agriculture. Early disease detection in plants, such as tomato plant that are susceptible to diseases, is critical because it helps farmers reduce losses and control the disease spread more effectively. However, the ability of machine to recognize diseased leaf objects is also influenced by the quality of data. Data collected directly from the field typically yields lower accuracy due to challenges faced in machine interpretation. To address this challenge, we propose a two-stage detection architecture for identifying infected tomato plant classes, leveraging YOLOv5 to detect objects within the images obtained from the field. We use Inception-V3 for classifying objects into known classes. Additionally, we employ a combination of two dataset: PlantDocs which represent field data, and PlantVillage dataset which serves as a cleaner dataset. Our experimental results indicate that the use of YOLOv5 in handling data under actual field conditions can enhance model performance, although the accuracy value is moderate (62.50 %)