IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
A merchant analytics framework for revenue forecasting and financial stress detection using transaction data
Get PDFBy processing payments and providing specialized financial services, acquiring banks are essential for merchants’ operations. To forecast 30-day revenue trajectories, identify seasonal demand patterns, and identify early indicators of financial stress, this paper presents a scalable merchant analytics framework that benefits from transactional data. The framework captures multi-level seasonalities using Prophet time series model, allowing dynamic product offerings like revenue-based loans. Proactive risk management is supported offerings like revenue-based loans. Proactive risk management is supported. by a new stress-flagging mechanism that identifies merchants at risk based on deviations in revenue trends. The framework achieved a median 30-day mean absolute percentage error (MAPE) of 56.51% after the validation on a dataset with 130,350 transactions from 460 merchants in a volatile economic environment. The model demonstrated significant practical utility in identifying financial distress and segmenting merchant behavior, despite its moderate predictive precision, which is common challenge in high-variance merchant datasets. Model outputs are converted into decision-support visualizations along with an interactive dashboard
Transformer and text augmentation for tourism aspect-based sentiment analysis
Get PDFThe 36.98% growth in the quantity of electronic word of mouth (e-WOM) over the past five years presents opportunities for the tourism industry to understand tourists' needs and desires better when analyzed effectively. Aspect-based sentiment analysis (ABSA) is proposed as a solution, as it can identify the sentiment at a more detailed aspect level. Prior research revealed two issues in ABSA: imbalanced datasets and poor performance in representing implicit aspects and opinions. The authors proposed a combination of the bidirectional and auto-regressive transformer (BART) and bidirectional encoder representations from transformers (BERT) models. Leveraging BART capability in modeling context and BERT expertise in modeling text semantics and nuances, the author proposed an ABSA model that combines BART and BERT using the ensemble method. The experimental results reveal that combining these models significantly enhances the performance of the ABSA model, with an F1-score reaching 70%. Furthermore, text augmentation and preprocessing did not bring improvements in ABSA performance
Personalized virtual reality therapy for children with autism spectrum disorder
Get PDFThe treatment of autism spectrum disorders (ASD) has often relied on broad therapeutic approaches that may not meet each individual's specific needs. This research highlights the importance of personalized therapy to address the unique sensory and emotional requirements of autistic children. We explore recent advances in therapeutic technologies, focusing on serious games and virtual reality (VR) as promising tools in this field. Our proposed solution is a VR application designed to provide a personalized, relaxing experience for children with autism. The application is tailored to accommodate individual preferences and sensory sensitivities, adjusting visual and auditory stimuli to reduce sensory overload and promote emotional regulation. This personalized approach aims to help children manage anxiety and stress more effectively
Broiler meats tenderness prediction using near infrared spectroscopy against non-linear predictive modelling
Get PDFNear infrared (NIR) spectroscopy is a non-invasive analytical technique known for its ability to assess the quality attributes of meat products. However, the linear models utilized, partial least square (PLS) and principal component regression (PCR) achieved unsatisfactory performances of meat physical attributes prediction. Hence, in this research, for its inherent advantages in modelling nonlinear system, artificial neural network (ANN) is augmented to the components of PCR and PLS. Through the augmentation, the principal component neural network (PCNN) and latent variable neural network (LVNN) models are developed. From the results obtained, it shows that PCNN and LVNN successfully surpassed their respective linear versions of PCR and PLS by 70% higher shear force prediction performances. The LVNN proved to achieve the best prediction in breast meat with root mean square error of prediction (RMSEP) of 0.0769 kg and coefficient of determination (RP2) of 0.8201 whilst for drumsticks, RMSEP=0.1494 kg and RP2=0.8606. NIR spectroscopy technology integrated with machine learning yields a promising non-invasive technique in predicting the shear force of intact raw broiler meat
Enhancing communication and interaction in the movie industry based SparkMLlib's recommendation system
Get PDFIn the ever-evolving landscape of streaming platforms, recommendation systems contribute significantly to enhancing the user experience. This article examines the significance of these systems in suggesting movies, analyzing their impact on user satisfaction and platform performance. Utilizing SparkMLlib, a powerful tool for large-scale data processing, we explore various recommendation techniques, including collaborative filtering and content-based filtering. We highlight the dimension of digital communication to further enhance the accuracy of recommendations and foster greater user engagement. Our study also addresses the challenges and future opportunities related to recommendation systems, emphasizing the need for transparency and ethical algorithms. This research highlights the potential for recommendation systems to revolutionize the digital entertainment landscape and shape the future of the movie industry
Deep learning approaches for Braille detection and classification: comparative analysis
Get PDFThis study proposes a hybrid approach to Braille translation leveraging the strengths of both YOLO for object detection and multitude of classification models such as ResNet, and ResNet for accurate Braille character classification from images. Upon comparing numerous models on various performance metrics, ResNet and DenseNet outperformed other models, exhibiting high accuracy (0.9487 and 0.9647 respectively) and F1-scores (0.9481 and 0.9666) due to their deep, densely connected architectures adept at capturing intricate Braille patterns. CNNs with pooling showed balanced results, while MobileNetV2's lightweight design limited complex classification. ResNeXt's multi-path learning achieved respectable performance but lagged behind ResNet and DenseNet. In the future the results from our study could be further explored on contracted Braille recognition, be adapted to various Braille codes, and optimized for mobile devices, for real time Braille detection and translation on smartphones
Exploring social media sentiment patterns for improved cyberbullying detection
Get PDFCases of online bullying and aggressive behaviors directed at social media users have surged in recent years. These behaviors have had negative impacts on victims from a wide range of demographic groups. While efforts have been made to address persistent digital harassment, the expected outcome has been limited due to the lack of effective tools to quickly identify cyberbullying behaviors and censor them accordingly on social media platforms. This study presents a scalable and systematic method to detect and analyze offensive behavior and bullying on Twitter (now known as X). Our methodology involves extracting textual, user-related, and network-related attributes to understand the traits of individuals involved in such behaviors. This approach aims to recognize distinctive characteristics that set them apart from regular users. This study proposes a novel model by employing an integrated deep-learning model, combining the bidirectional gated recurrent unit (BiGRU), transformer block, and convolutional neural network (CNN). This model aims to classify X comments into offensive and non-offensive categories. The proposed model’s efficiacy has been evaluated through several experiments by combining three widely recognized datasets of hate speech. The proposed model achieves an accuracy rate of approximately 98.95%, showing promising results in identifying and categorizing offensive behavior in cyberbullying
Child-friendly e-learning for artificial intelligence education in Indonesia: conceptual design
Get PDFDue to the widespread use of smartphones, most children in Indonesia are now engaged in playing video games. To make these games more exciting and challenging, video game manufacturers often incorporate artificial intelligence (AI). While various studies have highlighted the benefits of playing video games for children, this research has revealed some significant negative impacts that need to be addressed, as they can affect children's prospects. One of the major detrimental effects is the growing negative perception towards robots and AI, with concerns that they will replace human jobs. To counteract these negative impacts, educational institutions in Indonesia need to proactively plan and prepare for the consequences of gaming through formal learning. Given Indonesia's vast territory, consisting of islands, and its large population, it is crucial to implement appropriate learning technology. This article presents the architectural design of a child-friendly e-learning system that focuses on teaching children about AI. The design considers the available technology in Indonesia, based on our experience. The child-friendly e-learning model for AI education is expected to cultivate an interest in learning about technology, thus diverting children's attention from video game addiction
Intelligent route optimization for internet of vehicles using federated learning: promoting green and sustainable IoT networks
Get PDFAs the internet of vehicles (IoV) continues to evolve, optimizing vehicle routing becomes increasingly important for enhancing traffic efficiency and minimizing environmental impact. This paper introduces an intelligent vehicle route optimization protocol leveraging federated learning (FL) to achieve green and sustainable IoV systems. By distributing the learning process across multiple edge devices, the proposed protocol minimizes the need for centralized data processing, reducing network congestion, and preserving user privacy. The system optimizes vehicle routes based on real time traffic conditions, fuel efficiency, and carbon emissions, and promoting greener transportation practices. Simulations conducted in a dynamic IoV environment demonstrate significant improvements in route efficiency, fuel consumption, and carbon emissions. The results underscore the potential of FL in transforming IoV routing by balancing performance and sustainability, making it a promising solution for the future of connected transportation
Classification system of banana types and ripeness levels based on convolutional neural network
Get PDFRecently, the availability of bananas in supermarkets has been relatively abundant. However, most buyers experience problems categorizing the type and level of ripeness of bananas, so the level of purchases of this fruit decreases. This study implements a system that can automatically classify bananas based on type and level of ripeness so that buyers can choose them based on their needs. In this study, the proposed system could classify the types and degrees of banana ripeness using a Convolutional Neural Network (CNN) where the system was implemented in real-time using the hardware of the Jetson Nano as a processing unit and a camera system as a sensor. The methodology adopted in this research involves implementing CNN architectures, i.e., ResNet-18 and ResNet-50, under various conditions. The training phase comprises 60 epochs, while testing considers illumination parameters from LED lights with power of 6 watts, 12 watts, and 22 watts under distances ranging from 10 to 100 cm. The results show that the system could classify the type and level of ripeness of bananas in real-time with an accuracy of 93% that is achieved using the 22-watt power for all type and ripeness levels.