Journal of Ambient Intelligence and Humanized Computing最新文献

This paper is the first of a two-part study aiming at building a low-cost visible-light eye tracker (ET) for people with amyotrophic lateral sclerosis (ALS). The whole study comprises several phases: (1) analysis of the scientific literature, (2) selection of the studies that better fit the main goal, (3) building the ET, and (4) testing with final users. This document basically contains the two first phases, in which more than 500 studies, from different scientific databases (IEEE Xplore, Scopus, SpringerLink, etc.), fulfilled the inclusion criteria, and were analyzed following the guidelines of a scoping review. Two researchers screened the searching results and selected 44 studies (-value = 0.86, Kappa Statistic). Three main methods (appearance-, feature- or model- based) were identified for visible-light ETs, but none significantly outperformed the others according to the reported accuracy -p = 0.14, Kruskal–Wallis test (KW)-. The feature-based method is abundant in the literature, although the number of appearance-based studies is increasing due to the use of deep learning techniques. Head movements worsen the accuracy in ETs, and only a very few numbers of studies considered the use of algorithms to correct the head pose. Even though head movements seem not to be a big issue for people with ALS, some slight head movements might be enough to worsen the ET accuracy. For this reason, only studies that did not constrain the head movements with a chinrest were considered. Five studies fulfilled the selection criteria with accuracies less than (2^{circ }), and one of them is illuminance invariant.

With the rapid development and wide proliferation of sensor devices and the Internet of Things (IoT), machine learning algorithms processing and analysing one or more modalities of sensory signals have become an active research field given its numerous applications, particularly in the domestic intelligent environment (DIE). In the past decades, the research on sensing and interactive devices of DIE and deep learning (DL) based methods have become strikingly popular. Several missions, such as the pro- cessing and analysis of sensing signals related to domestic instruments and the control of certain devices to act upon the results, comprise the main working targets in DIE. The goal of this review is to provide a brief overview of the aforementioned sensors, their related DL algorithms and their applications. To comprehend the ideas behind the use of various devices found in domestic intelligent instruments, we first summarize the available information. Then, to quantify and adapt the residents’ knowledge of the household environment, we review data-driven learning techniques based on the aforementioned sensor-based devices and introduce robotic applications that provide helpers and action outputs in the environment. Finally, we investigate the commonly utilized datasets relevant to DIE and human activ- ity recognition (HAR) and explore the challenges and prospects of their applications in the DIE field.

This paper presents a rich class of information theoretical measures designed to enhance the accuracy of portfolio risk assessments. The Mean-Variance model, pioneered by Harry Markowitz, revolutionized the financial sector as the first formal mathematical method to risk-averse investing in portfolio optimization theory. We analyze the effectiveness of this with the models that replace expected portfolio variance with measures of information (uncertainty of the portfolio allocations to the different assets) and five major practical issues. The empirical analysis is carried out on the historical data of Indian financial stock indices by application of portfolio optimization problem with information measures as the objective function and constraints derived from the return and the risk. Our findings indicate that the information measures with parameters can be used as an adequate supplement to traditional portfolio optimization models such as the mean-variance model.

In this editorial, we explore the urgent challenges created by the rise of infodemics —a term used to describe the epidemic spread of fake news, misinformation, and disinformation through social networks initially associated with the COVID-19 pandemic. This issue has drawn significant attention from various academic fields, including computer science, artificial intelligence, mathematics, physics, biology, sociology, and psychology, among others. This special issue is dedicated to advancing infodemics research across various academic domains. The selected papers include relevant contributions advancing the state of the art in the area, ranging from network analysis for identifying influential nodes and communities in networks to language processing for text classification and filtering relevant messages within extensive corpora.

Representing fitness-related data physically can better help people gain awareness and reflect on their physical activity behavior. However, there has been limited research conducted on the impact of physicalizing personal data in a public context, particularly regarding its effect on motivations for physical activity. Augmenting the physical environment with interactive technology holds great promise in facilitating outdoor physical activity. To explore the design space of data physicalization-based interactive environments, we created TIDAL, a design concept that provides physical rewards in the form of tiles on the road to acknowledge runners’ goal achievements. We created a video prototype as a probe to gather insights through semi-structured interviews with six recreational runners to evaluate TIDAL. The co-constructing stories method, a participatory design technique, was employed during these interviews to facilitate qualitative evaluation. The results of our study showed that TIDAL has the potential to increase runners’ motivation. We reported the key insights derived from participants’ feedback and co-constructed stories and discussed the broader implications of our work.

Abstract

The research field of this paper is unsupervised learning in machine learning, aiming to address the problem of how to simultaneously resist feature attacks and improve model classification performance in unsupervised learning. For this purpose, this paper proposes a method to add an optimized loss function after the graph encoding and representation stage. When the samples are relatively balanced, we choose the cross-entropy loss function for classification. When difficult-to-classify samples appear, an optimized Focal Loss*() function is used to adjust the weights of these samples, to solve the problem of imbalanced positive and negative samples during training. The developed method achieved superior performance accuracy with the values of 0.721 on the Cora dataset, 0.598 on the Citeseer dataset,0.862 on the Polblogs dataset. Moreover, the testing accuracy value achieved by optimized model is 0.745, 0.627, 0.892 on the three benchmark datasets, respectively. Experimental results show that the proposed method effectively improves the robustness of adversarial training models in downstream tasks and reduces potential interference with original data. All the test results are validated with the k-fold cross validation method in order to make an assessment of the generalizability of these results.

This study addresses the crucial task of accurately classifying brainwave signals associated with distinct brain states, utilizing five supervised machine learning algorithms: K-Nearest Neighbors, Support Vector Machine, Decision Tree, Linear Discriminant Analysis, and Logistics Regression. The primary objectives encompass developing and optimizing these models, assessing the impact of hyperparameter tuning on performance through metrics like accuracy, consistency, and prediction time, and creating a user-friendly web-based deployment interface. The Decision Tree model emerges with the highest average accuracy score of 90.03%, swift prediction times, and notable consistency. Following hyperparameter tuning, SVM and LR showcase substantial accuracy gains (15.63% and 1.50% respectively), enhancing all models' consistency. KNN and SVM are identified as the top-performing algorithms for accurate brain state classification. This study's findings hold implications for neuroscience research, human–computer interaction, healthcare diagnostics, and assistive technologies, offering insights into both effective algorithm selection and the role of hyperparameter tuning in optimizing model performance.

Demands for cost-efficient and just-in-time service systems have rapidly increased due to the present-day competitive resource allocation. We focus on optimizing policies for highly efficient service systems because customer congestion often arises from suboptimal policies rather than flawed arrangements. Quasi and metaheuristic optimization techniques are widely employed to establish cost-optimal service policies, mitigating customer congestion, primarily caused by unplanned policies or inadequate facilities. This article initially introduces a notion of unreliable service and the F-policy for stochastic modeling of finite capacity customer service systems. Next, we utilize the recently-developed and proficient Grey Wolf Optimizer, a metaheuristic approach, along with the Quasi-Newton method, to determine the optimal values of decision parameters for a cost-efficient service systems. This is achieved through extensive numerical experiments that encompass diverse service characteristics, customer behavior, and performability measures. The results emphasizes the importance of both preventive and corrective actions for enhancing service system efficiency. Our findings also highlight the practicality of the Grey Wolf Optimization approach and stochastic modeling in achieving efficient policies and optimizing performance for the studied service model. In general, the F-policy is widely adopted for controlling queueing systems across various industries such as telecommunications, transportation, and healthcare, where maintaining reasonable wait times, service levels, and system stability is crucial. This article contributes to the mathematical modeling of this approach. Nonetheless, further research is needed to validate and simulate these findings in industrial settings.

Addressing the inherent hazards of on-site construction work and stagnant labor productivity is crucial in the construction industry. To tackle these challenges, automated monitoring of construction sites and analysis of workers' actions play a pivotal role. In this study, we developed a method for classifying actions at a construction site from video, using deep learning. Specifically, we used two image processing techniques, pose assessment and object detection, and found that the accuracy of action classification was improved by extracting information on the proximity of workers to equipment installed at the construction site, and also by considering the pose information. For classification, LSTM (Long Short-Term Memory), CNN (Convolutional Neural Network), and XGBoost models were used, and the presence of proximity information improved average recall by 7.0% to 8.5% for all models used. The final model was developed as an ensemble of these methods, offering accuracy and average recall that are higher than with conventional methods. The methodology developed in this research enables quantification and visualization of work content at construction sites, contributing to the overall enhancement of safety and productivity within the construction industry.

This research proposes an Intelligent Decision Support System for Ground-Based Air Defense (GBAD) environments, which consist of Defended Assets (DA) on the ground that require protection from enemy aerial threats. A Fire Control Officer is responsible for assessing threats and assigning the most appropriate weapon to neutralize them. However, the decision-making process can be prone to errors, risking resource wastage and endangering DA protection. To address this problem, this research proposes a hybrid approach that combines a knowledge-driven fuzzy inference system with machine learning models to optimize resource allocation while incorporating expert knowledge in the decision-making process. Since sensory data obtained from multiple radars may be incomplete or incorrect, a fuzzy knowledge graph-based system is used for data fusion and providing it to the connected modules. Feature selection is optimized by including the most important parameters, such as the vitality of defended assets and threat score, in the threat evaluation. The results from these subsystems are visualized using a Geographical Information System, allowing for real-time mapping of the GBAD environment and displaying the results in a user-friendly web interface. The proposed system has undergone rigorous testing and evaluation, resulting in an efficient and accurate weapon assignment model with a low RMSE value of 0.037. Overall, this Intelligent Decision Support System provides an effective solution for optimizing decision-making processes in GBAD environments and can significantly improve DA protection.