Information Dimension Research Articles

Deep learning techniques to process 3D chest CT

The idea of using X–rays and Computed Tomography (CT) images as diagnostic method has been explored in several studies. Most of these studies work with slices of CT image in 2D, requiring less computational capacity and less time to process them than 3D. The processing of volumetric data (the complete CT images in 3D) adds an extra dimension of information. However, the magnitude of the data is considerably larger than working with slices in 2D, so extra computational processing is required. In this study a model capable of performing a classification of a 3D input that represents the volume of the CT scan is proposed. The model is able to classify the 3D input between COVID–19 and Non–COVID–19, but reducing the use of resources when performing the classification. The proposed model is the ResNet–50 model with a new dimension of information added, which is a simple autoencoder. This autoencoder is trained on the same dataset, and a vector representation of each exam is generated and used together with the exams to feed the ResNet–50. To validate the proposal, the same proposed model is compared with and without the autoencoder module that provides more information to the proposed model. The proposed model obtains better metrics than the same model without the autoencoder, confirming that extracting relevant features from the dataset helps improve the performance of the model.

Editorial

Dear Readers, It gives me great pleasure to announce the sixth regular issue of 2024. In this issue, 6 papers cover various topical aspects of computer science by 19 authors from 7 countries: Brazil, Chile, France, India, Saudi Arabia, Tunisia, and Turkey. In an ongoing effort to further strengthen our journal, I would like to expand the editorial board: If you are a tenured associate professor or above with a strong publication record, you are welcome to apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and trends. As always, I would like to thank all authors for their sound research and the editorial board and our guest reviewers for their extremely valuable review effort and suggestions for improvement. These contributions, together with the generous support of the consortium members, help to maintain the quality of our journal.  In this regular issue, I am very pleased to introduce the following 6 accepted articles: George Marsicano, Edna Dias Canedo, Glauco V. Pedrosa, Cristiane S. Ramos, and Rejane M. da C. Figueiredo from Brazil look in their study into digital transformation of public services in a startup-based environment by 23 focus groups and 175 participants in total. Mauricio Solar and Pablo Aguirre from Chile discuss their research on 3D chest CT processes applying a ResNet-50 model to which a new dimension of information has been added, namely a simple autoencoder. In a collaborative work between researchers from Tunisia and France, Rakia Saidi, Fethi Jarray, and Didier Schwab propose in their article a cross-encoder neural network (Cross-BERT-GRU) to deal with the semantic similarity of Arabic sentences that benefits from both the strong contextual understanding of BERT and the sequential modeling capabilities of GRU. Also in a collaborative research between institutions from Tunisia and Saudi Arabia, Nozha Jlidi, Sameh Kouni, Olfa Jemai, and Tahani Bouchrika present their research on MediaPipe with GNN for human activity recognition. G.V.Vidya Lakshmi and S. Gopikrishnan from India look into missing values research for the IoT domain and in particular present IMD-MP technique that improves imputation accuracy for big data analysis in IoT applications based on spatial-temporal correlations. Last but not least, F. Didem Alay, Nagehan İlhan, and M. Tahir Güllüoğlu address in their article a comparative study of data mining methods for solar radiation and temperature forecasting models. Enjoy Reading Cordially,  Christian Gütl, Managing Editor Graz University of Technology, Graz, Austria

Beyond mutations: Accounting for quantitative changes in the analysis of protein evolution

Molecular phylogenetic research has relied on the analysis of the coding sequences by genes or of the amino acid sequences by the encoded proteins. Enumerating the numbers of mismatches, being indicators of mutation, has been central to pertinent algorithms. Specific amino acids possess quantifiable characteristics that enable the conversion from “words” (strings of letters denoting amino acids or bases) to “waves” (strings of quantitative values representing the physico-chemical properties) or to matrices (coordinates representing the positions in a comprehensive property space). The application of such numerical representations to evolutionary analysis takes into account not only the occurrence of mutations but also their properties as influences that drive speciation, because selective pressures favor certain mutations over others, and this predilection is represented in the characteristics of the incorporated amino acids (it is not born out solely by the mismatches). Besides being more discriminating sources for tree-generating algorithms than match/mismatch, the number strings can be examined for overall similarity with average mutual information, autocorrelation, and fractal dimension. Bivariate wavelet analysis aids in distinguishing hypermutable versus conserved domains of the protein. The matrix depiction is readily subjected to comparisons of distances, and it allows the generation of heat maps or graphs. This analysis preserves the accepted taxa order where tree construction with standard approaches yields conflicting results (for the protein S100A6). It also aids hypothesis generation about the origin of mitochondrial proteins. These analytical algorithms have been automated in R and are applicable to various processes that are describable in matrix format.

Entity type inference based on path walking and inter-types relationships

As a crucial task for knowledge graphs (KGs), knowledge graph entity type inference (KGET) has garnered increasing attention in recent years. However, recent methods overlook the long-distance information pertaining to entities and the inter-types relationships. The neglect of long-distance information results in the omission of crucial entity relationships and neighbors, consequently leading to the loss of path information associated with missing types. To address this, a path-walking strategy is utilized to identify two-hop triplet paths of the crucial entity for encoding long-distance entity information. Moreover, the absence of inter-types relationships can lead to the loss of the neighborhood information of types, such as co-occurrence information. To ensure a comprehensive understanding of inter-types relationships, we consider interactions not only with the types of single entity but also with different types of entities. Finally, in order to comprehensively represent entities for missing types, considering both the dimensions of path information and neighborhood information, we propose an entity type inference model based on path walking and inter-types relationships, denoted as “ET-PT”. This model effectively extracts comprehensive entity information, thereby obtaining the most complete semantic representation of entities. The experimental results on publicly available datasets demonstrate that the proposed method outperforms state-of-the-art approaches.

Full Stokes Mid-Wavelength Infrared Polarization Photodetector Based on the Chiral Dielectric Metasurface

Conventional imaging techniques can only record the intensity of light while polarization imaging can record the polarization of light, thus obtaining a higher dimension of image information. We use the COMSOL software to numerically propose a circular polarization photodetector composed of the dislocated 2-hole Si chiral metasurfaces controlling the circular polarization lights and the HgCdTe (MCT) photodetector chip to detect the intensity of light signals. The chiral metasurfaces can be equated to a significant radiation source of the Z-type current density under the right circularly polarized incidence conditions, which explains the large circular dichroism (CD) of absorption of 95% in chiral photodetectors. In addition, the linear dichroism (LD) of the linear polarization pixel is 0.62, and the extinction ratio (ER) is 21 dB. The full Stokes pixel using the six-image-element technique can almost measure arbitrary polarization information of light at 4 μm operation wavelength. Our results highlight the potential of circular dichroic metasurfaces as photonic manipulation platforms for miniaturized polarization detectors.

A threat assessment method based on correlation-similarity information and three-way decisions in interval intuitionistic fuzzy environment

Threat assessment of targets is a critical component of combat decision-making. Existing threat assessment methods only consider specific dimensions of information. When information is missing or abnormal, they may fail to provide accurate assessment results. Moreover, most methods can only obtain the threat level of targets without providing target threat levels based on the battlefield environment. Therefore, in the context of interval intuitionistic fuzziness, this paper proposes a threat assessment method based on Grey Correlation Analysis and Technique for Order Preference by Similarity to an Ideal Solution (GCA-TOPSIS) and three-way decisions. This method utilizes interval intuitionistic fuzzy entropy to calculate attribute weights. It simultaneously considers information on both similarity and correlation between targets. By introducing grey correlation coefficients to balance the similarity and correlation between targets and the ideal solution, the method computes threat degrees based on the information from both dimensions. Through fuzzy evaluation information, it constructs loss functions, calculates decision thresholds, and thereby determines the threat levels of the targets. This approach is better suited for the complex and dynamic battlefield environment, providing decision-makers with target threat levels. Finally, we conducted simulation experiments under both complete data and data anomaly conditions. The results indicate that when the data noise reaches 0.5 dB, the absolute error of GCA-TOPSIS is only 0.08. When 66.67% of the data is missing, the error is 0.16. In both scenarios, the error is much smaller than that of other threat assessment methods. This demonstrates the method's strong robustness and its suitability for complex battlefields.

Time-varying characteristics of acoustic emission and fractals based on information dimension during structural failure of coal subjected to uniaxial compression

Uniaxial compression experiments were conducted on coal samples with different initial damage degrees, and the acoustic emission (AE) signals of the whole deformation and destruction process were captured synchronously. The distribution patterns of the acoustic emission signals at different loading stages were discussed. The chaotic system was introduced to process the acoustic emission series. The results indicate that the acoustic emission responses of coal samples with different damage levels had obvious stage characteristics during the whole destruction process. According to the information dimension fitting curves, the coal samples had good fractal characteristics in the whole loading process, and the value of information dimension decreased with the increased of threshold value.The information dimension of the acoustic emission energy signals showed a dynamic evolution law of “increase-decrease-increase” with time, and the rapid decreased of the information dimension could be used as a stable and effective precursor for predicting the macro-damage of coal.

The Commander’s Place on the Battlefield

News of Russian generals getting killed on the battlefields raises the question of whether a commander should lead an operation remotely, taking advantage of improved technical capabilities, or whether his presence close to the battlefield is necessary to meet the needs of his troops. The requirements of multi-domain operations and multi-domain integration call for command posts with enhanced capabilities along human, informational and physical dimensions. Unfortunately, these types of command posts have become very large, easily recognisable and vulnerable to attack. Alexander Kohli argues that commanders should not exclusively remain in remote command posts but should also provide support close to the front through personal contact with subordinate commanders and troops. Therefore, they must be able to rely on a command post concept encompassing several smaller forward and alternate command posts operating inconspicuously, yet capable of using a comprehensive command and control system. ◼

Knowledge, attitudes, and practices of human papillomavirus and self-sampling among adult women: a cross-sectional study.

This study aimed to investigate the knowledge, attitude, and practice (KAP) of human papillomavirus (HPV) and self-sampling among adult women. The cross-sectional, questionnaire-based study included adult women at Shanghai Pudong Hospital from October 14, 2022, to March 31, 2023. The questionnaire contained demographic information, knowledge, attitude and practice dimensions. Factors associated with KAP and self-sampling were identified by multivariate logistic regression. A total of 1843 valid questionnaires were collected. The average knowledge, attitude, and practice score was 10.09 ± 5.60, 26.76 ± 3.80, and 6.24 ± 2.20, respectively. Urban residents (estimate = 0.705, p < 0.001), suburban residents (estimate = 0.512, p < 0.001), as well as individuals with undergraduate degrees and higher (estimate = 0.535, p < 0.001), were associated with good knowledge, while individuals lacking a history of HPV infection (estimate = -0.461, p < 0.001) and married individuals (estimate = -0.185, p < 0.001) were less likely to have good knowledge. Higher knowledge scores (estimate = 0.087, p < 0.001) and individuals with undergraduate education and above (estimate = 1.570, p < 0.001) were associated with a positive attitude. Being married (estimate = 0.291, p = 0.049) was associated with good practice, whereas not engaging in sexual activity (estimate = -0.959, p < 0.001) or lacking a history of HPV infection (estimate = -0.499, p = 0.011) were associated with unfavorable practices. Minorities (OR = 2.787, p = 0.038) and individuals with multiple sexual partners (OR = 2.297 for two partners, OR = 2.767 for three or more partners, p = 0.020 and p = 0.022) were positively associated with self-sampling. However, higher knowledge (OR = 0.952, p = 0.026) and attitude scores (OR = 0.929, p = 0.015) were negatively associated with self-sampling. Demographic and behavioral factors significantly influenced KAP scores and self-sampling behaviors regarding HPV. Urban residency, higher education levels, positive attitudes, and minority status correlated with favorable outcomes, while factors like marriage and lack of sexual activity were associated with less favorable practices.

CTITF: A tensor factorization model with constrained bidirectional user trust and implicit feedback for context-aware recommender systems

Recommender systems offer an efficient solution to the problem of information overload, which is exacerbated by the rapid expansion of data. Context-aware recommender systems (CARS) have become prominent, incorporating contextual information to provide more desirable recommendations. While the dimension of contextual information rises, the complexity of algorithms increases as well, highlighting the data sparsity problem. In this paper, we propose CTITF, a novel tensor factorization model that incorporates constrained user bidirectional trust and implicit feedback. First, CTITF distinguishes the roles of trust networks in terms of the directionality of trust relationships: the trustor and the trustee. CTITF then constructs trust influences and biases for both roles to linearly represent user preferences. Furthermore, CTITF improves the rating prediction function and recommendation objective function based on CP factorization. Finally, CTITF filters out trust relationships where users do not have the same history items. It also uses contextual information to limit the amount of implicit feedback, addressing the issue of high time complexity. Experiments on two real-world datasets illustrate that the proposed model outperforms previous TF-based models and social recommender systems in terms of rating prediction accuracy. In addition, the CTITF model reduces the training time by 50.28% compared to the trustTF model, which also uses trust.

When Science Meets Digital Branding

The science media ecosystem continues to dramatically change with the proliferation of digital media channels and platforms. While there are many different types of digital media filling this void, the use of online videos, particularly on YouTube, has become one of the main sources for information, specifically for science and environmental issues. While most researchers examine the content of online science news and information, we show that digital trace data, including hyperlinks, “likes”, and comments that are unique to participatory online media, can reveal interesting patterns about how and why people communicate and engage with content as well as other community members. Using climate change as a case study, we demonstrate how researchers can harness hyperlinks to understand who produces controversial science news and information online, and the motivations underlying how content creators use of hyperlinks for science communication and public engagement. While many existing studies in science communication have focused on understanding the political dimension of online science news and information, our paper reveals that digital branding is a crucial part of science communication on video-based platforms such as YouTube and TikTok. We find that amateur content creators and new media outlets use hyperlinks to promote their channel and other social media presences to raise revenue. While self-promotion through citing links to one’s other social media platforms is associated with more views, linking to revenue raising websites is not associated with more views. We discuss the implications of these contexts for the online science news and information ecosystem.

Information Dimension Matching in Memristive Computing System for Analog Deployment of Deep Neural Networks

AbstractMemristor, with the ability of analog computing, is widely investigated for improving the computing efficiency of deep neural networks (DNNs) deployment. However, how to fully take advantage of the analog computing ability of memristive computing system (MCS) for DNN deployment is still an open question. Here, a new neural network models deployment scheme, that is, an information dimension matching (IDM) scheme, is proposed to fully take advantage of the analog computing ability of MCS. Furthermore, the spatial and temporal DNN, that is convolutional neural network (CNN) and recurrent neural network (RNN) is used to verify the proposed deployment scheme, respectively. The experimental results indicate that, compared to the traditional deployment schemes, the proposed deployment scheme shows obvious inference accuracy and energy efficiency improvement (&gt;4 × in four‐layer DNNs deployment), and the energy efficiency improvement increases dramatically with the layers increment of DNNs. This work paves the path for developing high computing efficiency analog MCS.

A Dynamic Detection Method for Railway Slope Falling Rocks Based on the Gaussian Mixture Model Segmentation Algorithm

Rockfall intrusion detection is crucial for the safety management of railway operations, and video detection methods help reduce deployment costs and improve detection efficiency. Mainstream neural network-based video detection methods have rapidly evolved in recent years but struggle to adapt to complex scenarios such as existing railway slope constructions due to weak generalization ability, low accuracy, and limited information acquisition. Therefore, this paper introduces a dynamic neural network detection model and establishes a dataset for rockfall intrusions in existing railway slope scenarios. The model initially relies on the YOLOv5 neural network, adopting an activation function suitable for the target scenario and addressing overfitting to achieve precise target recognition. Based on the neural network, the model dynamically detects rolling rockfalls by integrating a background subtraction algorithm based on the Gaussian Mixture Model and captures target dimensions using monocular vision technology, thus broadening the dimensions of detection information. Trials conducted on a railway in Shandong, China, demonstrate that the model accurately identifies moving rockfalls along the railway slopes and acquires the dimensions of moving rockfalls, successfully filtering out low-risk targets in the scene.

CONTEXTS.py (CS.py): A supervised contextual post-classification method to access multiple dimensions of complex geospatial objects

The qualitative dimensions of visible features in space can be captured by connecting spatial configurations arranged in a variety of different ways to diverse conceptual spaces. By conceptual spaces, we intend mental concepts describing specific spatial configurations present in a geographical area, defined by the contextual relationships among their constitutive elements.This paper presents a new supervised post-classification method allowing the extraction of semantically complex spatial objects from a single image of the Earth as, for instance, diverse conceptual spaces referring to multiple dimensions of land use (temporal, cultural, social, etc.).Computationally, our method is operationalised by CONTEXTS.py (CS.py), a plugin written in Python for QGIS. CS.py relies on training areas, defined by the user at diverse scales, to identify and extract in the input image conceptual spaces whose spatial contexts have the same spatial features present in the training areas.Applied to a case study on the island of Sicily, where millennial land use dynamics have resulted in a mosaic landscape, CS.py could detect from an orthophoto diverse conceptual spaces of land use in an area ordinarily classified as one land cover, thus expanding the capabilities of geospatial analysis to reach additional qualitative dimensions of information from image data.•CS.py simplifies a supervised contextual post-classification routine in an easy-to-use, practical and accessible QGIS plugin;•CS.py joins a family of tools for supervised object-based classification (e.g. OTB, GRASS), providing, additionally, the possibility to include contextual information as spatial criteria to train the classification routine.•CS.py has broad applications in different disciplines investigating landscape from quantitative and qualitative perspectives, allowing both, as in multiple environments.

Civilian-Military Integration in Ukrainian Defense Supply Chain

This research delves into the intricacies of the Ukrainian military’s supply chain during the ongoing Russian-Ukrainian war, with a specific focus on the significant role played by the civilian population in supporting the armed forces. Termed a defense supply chain, our study employs a multilevel model encompassing logistical, informational, and financial dimensions to elucidate the multifaceted levels within the defense supply chain network. Key actors identified include manufacturers, non-governmental organizations (NGOs), donors, and conscripts. The study sheds light on the vulnerabilities in the flow of goods in the defense supply chain. A resilient defense supply chain relies on efficient NGO communication and the crucial alignment of donor income growth with escalating resource prices.

Spatial pattern analysis reveals intra-specific competition and fractal characteristics of Pinus sylvestris var. mongolica populations in Xiaotaojia Gulley Watershed

The imprudent cultivation of Pinus sylvestris var. mongolica has led to degradation in the watersheds surrounding Chongli County. This study investigates the artificial forests of P. sylvestris var. mongolica in the Xiaotaojiagou watershed, Chongli County. Twenty-five subplots measuring 20 m × 20 m were established at upper, upper-middle, middle, and lower positions within the watershed. Spatial distribution patterns and fractal characteristics of P. sylvestris var. mongolica were analyzed using univariate pair correlation function g(r), marked correlation functions (marked correlation function Kmm(r), marked variation function γm(r), and Schlather’s Im(r) function), and fractal dimensions (box-counting dimension, information dimension). The findings are as follows: (1) P. sylvestris var. mongolica in the middle section of the watershed displayed the largest average diameters at breast height and tree height, while those in the lower section exhibited the largest average crown width. (2) P. sylvestris var. mongolica populations in the upstream, middle, and downstream sections of the watershed exhibited three alternating states: aggregated distribution, uniform distribution, and random distribution. However, with increasing spatial scale, the distribution gradually tended towards randomness. P. sylvestris var. mongolica in the upper-middle section of the watershed showed aggregated distribution at scales of 1.5–32.5 m, transitioning to random distribution and then to uniform distribution with increasing spatial scale. (3) Individual attributes of P. sylvestris var. mongolica (diameter at breast height, tree height, and crown width) exhibited high correlation and facilitation in the upstream and upper-middle sections at small-to-medium scales, whereas negative correlations and inhibitions were observed in the middle and lower sections. These attributes tended to be uncorrelated with increasing spatial scale. (4) Fractal dimensions indicated that the degree of aggregation of P. sylvestris var. mongolica population was lowest in the downstream section of the watershed. In conclusion, spatial distribution patterns and intraspecies relationships of P. sylvestris var. mongolica populations are influenced by spatial scale and environmental conditions, particularly in more resource-competitive downstream areas. Effective management and protection of P. sylvestris var. mongolica populations require targeted measures such as adjusting afforestation density, improving soil conditions, and implementing differentiated ecological restoration strategies to alleviate intraspecies competitive pressures and promote population health and sustainable development.

Dual-branch network with fused Mel features for logic-manipulated speech detection

In recent years, logic-manipulated speech detection has predominantly relied on single-feature and single-channel networks, which fail to consider a broader range of acoustic features. To address this limitation, we propose a novel approach based on a dual-branch network that incorporates fused Mel features for detecting logic-manipulated speech. The structure of our network consists of three main components. Firstly, we extract Mel spectrograms and Mel-frequency cepstral coefficients (MFCC) to address the deficiencies caused by the reduction in the information dimension. Then, to better utilize these two types of features, we construct a dual-branch network framework that enables the separate learning of temporal and frequency domain characteristics. Finally, at the backend of the network, we fuse the feature vectors for decision-making. Experimental evaluations conducted on the ASVspoof2019LA dataset demonstrate the effectiveness and superiority of our proposed method compared to state-of-the-art models and other acoustic features.

Prediction of dynamical systems from time-delayed measurements with self-intersections

In the context of predicting the behaviour of chaotic systems, Schroer, Sauer, Ott and Yorke conjectured in 1998 that if a dynamical system defined by a smooth diffeomorphism T of a Riemannian manifold X admits an attractor with a natural measure μ of information dimension smaller than k, then k time-delayed measurements of a one-dimensional observable h are generically sufficient for μ-almost sure prediction of future measurements of h. In a previous paper we established this conjecture in the setup of injective Lipschitz transformations T of a compact set X in Euclidean space with an ergodic T-invariant Borel probability measure μ. In this paper we prove the conjecture for all (also non-invertible) Lipschitz systems on compact sets with an arbitrary Borel probability measure, and establish an upper bound for the decay rate of the measure of the set of points where the prediction is subpar. This partially confirms a second conjecture by Schroer, Sauer, Ott and Yorke related to empirical prediction algorithms as well as algorithms estimating the dimension and number of required delayed measurements (the so-called embedding dimension) of an observed system. We also prove general time-delay prediction theorems for locally Lipschitz or Hölder systems on Borel sets in Euclidean space.

Relationship between organizational justice and citizenship behavior as perceived by staff nurses at a selected hospital

Background Organizational justice is a managerial practice that may create a positive and more effective citizenship behavior among staff nurses. Staff nurses perceive fairness from their managers and have more satisfaction, motivation, and desire to achieve extra work. Aim To examine the relationship between organizational justice and citizenship behavior as staff nurses perceive. Design A descriptive correlational research design was utilized to carry out this study. Sample A convenience sample (281) of staff nurses. Setting The study conducted in all units at New Kasr Einy Teaching Hospital except operating rooms. Tools Two adopted questionnaires were used. The first questionnaire was the organizational justice questionnaire. The second questionnaire was the organizational citizenship behavior questionnaire. Results Finding of the current research reported that the highest mean percentages (85.5 and 71.3%) were given to interactional justice and informational justice dimensions, while the lowest mean percentages (38.4 and 36.1%) were given for procedural justice and distributive justice. Also, as regards citizenship behaviors dimensions, the highest mean percentages (89.4, 90.7, and 87.0%, respectively) were given to the dimensions of altruism, courtesy, and conscientiousness, while the lowest mean percentages (67.8 and 61.6%) were given to sportsmanship and civic virtue dimension and there was no statistical correlation between organizational justice and organizational citizenship behavior (P=0.73). Conclusion The present study results concluded that there was no statistically significant relation between organizational justice and citizenship behavior. Recommendations Hospital administrators should ensure nurse managers adhere to rules and regulations to achieve justice. Nurse managers should conduct periodical meetings with staff nurses to discuss and solve work problems considering staff nurses’ concerns and interests.

The Effects of Live Streamer's Facial Attractiveness and Product Type on Consumer Purchase Intention: An Exploratory Study with Eye Tracking Technology.

As a booming branch of online retailing, live-streaming e-commerce can present abundant information dimensions and diverse forms of expression. Live-streaming e-commerce has enabled online retailers to interact with customers face-to-face, resulting in widespread instances of emotional and impulse buying behavior. Prior research in live-streaming e-commerce has suggested that live streamers' characteristics, especially the live streamer's face, can affect customers' purchase intentions. The present research used questionnaire surveys and an eye tracking experiment to investigate the impact of live streamer's facial attractiveness on consumer purchase intention for search-based and experience-based products. The questionnaire survey analyzed 309 valid questionnaires and revealed that attractive faces are the key influencing factor driving consumers' impulse purchase intentions. Moreover, consumers' emotional experience plays a partial mediating role in the process of live streamers' faces influencing purchase intention. The eye tracking experiment further explored the mechanism of a live streamer's facial attractiveness on consumers' purchase intentions of search-based products and experience-based products from the perspective of visual attention by analyzing 64 valid sets of data. The results showed that attractive faces attract more consumers' attention and, therefore, increase their purchase intention. Furthermore, there is a significant interaction between product type, the live streamer's facial attractiveness, and consumers' purchase intentions. In the case of unattractive live streamers, consumers are more likely to buy search-based products than experience-based products, while the purchase intention does not vary between search-based products and experience-based products in the case of attractive live streamers. The present study provides evidence for 'beauty premium' in live-streaming e-commerce and sheds light on the design of the match between live streamers and different types of products.