Cultural heritage represents the historical and cultural achievements of a nation, playing a vital role in studying human civilization and preserving national languages and scripts. This study utilizes virtual simulation technology to design a virtual pavilion for Chinese language and writing, employing image and text feature extraction algorithms for feature fusion and 3D modeling. The effectiveness of Chinese character extraction is validated through feature point matching, while the virtual exhibition’s impact is assessed via user experience scores. Results indicate that the proposed algorithm achieves accurate extraction with no misrecognition. User interest rankings highlight text images as the most influential factor, followed by visual imagery, pavilion experience, scene art, and language culture. Analysis of user feedback shows an average experience score exceeding 60 points, confirming the pavilion’s effectiveness in preserving and promoting Chinese language and writing culture.

In recent years, due to the adjustment of economic structure, the people’s living standard and the increase of leisure time, the sports industry has become a new economic growth point. This paper studies and analyzes the characteristics of the industry background and business background of the sports industry, explores the factors and internal driving force affecting the design of its business model, and fully analyzes the mechanism, functional role, and logical relationship of the elements for constructing the business model of the sports industry, and then explores the characteristics of the business style of the sports industry. From the perspective of knowledge state, using the reinforcement learning mechanism, the evolution process of the sports industry business model from the first stage to the fourth stage is described. Taking Company H as a research case, the process and economic effect of the transformation and upgrading of its business model through the reinforcement learning mechanism is analyzed and it is found that as of 2023 the company’s operating income has increased by 2.4 times through transformation and upgrading, and its net profit has increased by 125.57 percentage points compared to 2016. It further understands the role that the enhanced learning mechanism brings to the development of the sports industry, and expects to be able to provide a reference for the sports industry to carry out business model transformation in the future.

We initiate a study of the toughness of directed graphs by considering the natural generalization of that for ordinary graphs. After providing some general results, computations are completed for a few natural examples. Maximum possible toughness is also considered. Some open problems are  posed.

Let \(G\) and \(H\) be graphs and \(1\) be a positive number. An \(H\)-irregular labeling of \(G\) is an assignment of integers from \(1\) up to \(k\) to either vertices, edges, or both in \(G\) such that each sum of labels in a subgraph isomorphic to \(H\) are pairwise distinct. Moreover, a comb product of \(G\) and \(H\) is a construction of graph obtained by attaching several copies of \(H\) to each vertices of \(G\). Meanwhile, an edge comb product of \(G\) and \(H\) is an alternate construction where the copies of \(H\) is attached on edges of \(G\) instead. In this paper, we investigate the vertex, edge, and total \(H\)-irregular labeling of \(G\) where both \(G\) and \(H\) is either a comb product or an edge comb product of graphs.

This study applies Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) algorithms to classify five types of basketball footwork. SVM maps the training data into a high-dimensional space using nonlinear transformation and classifies it with support vectors and a hyperplane. Experimental analysis showed minimal differences in peak and trough values of footwork movements; therefore, only mean and standard deviation features were retained, resulting in 12 effective features. KNN experiments demonstrated that recognition accuracy varies with different K values. The highest accuracy (80.7%) was achieved when K = 5 with the selected features. The study also examined the physical characteristics of basketball players, analyzing height, weight, and other indicators. Statistical results showed no significant body shape differences between experimental and control groups (P > 0.05). A T-test on dribbling, shooting, and layup performance also revealed no significant differences between the groups (P > 0.05).

This study explores the employment competitiveness of computer science majors by integrating combinatorial mathematics into the evaluation process. Utilizing the Analytic Hierarchy Process (AHP) and the improved FKCM clustering algorithm, we construct a hierarchical model to assess the impact of entrepreneurial education, learning motivation, and investment on job competitiveness. Data from 314 participants were analyzed using combinatorial techniques to derive optimal weightings for each factor, ensuring the evaluation model’s robustness. The results highlight significant gender differences in practical and feedback-based entrepreneurship education, with males outperforming females. However, no notable differences were observed in job interest, learning motivation, or overall employment competitiveness.

An (unrooted) binary tree is a tree in which every internal vertex has degree \(3\). In this paper, we determine the minimum and maximum number of total dominating sets in binary trees of a given order. The corresponding extremal binary trees are characterized as well. The minimum is always attained by the binary caterpillar, while the binary trees that attain the maximum are only unique when the number of vertices is not divisible by~\(4\). Moreover, we obtain a lower bound on the number of total dominating sets for \(d\)-ary trees and characterize the extremal trees as well.

This paper proposes an optimized Backpropagation (BP) neural network for improving intelligent elderly care talent training. To address BP’s limitations, including noise sensitivity and slow convergence, we introduce Particle Swarm Optimization (PSO) to refine network weights and thresholds. The model integrates course quality, teacher effectiveness, platform support, and market demand, aiming to optimize elderly care service talent cultivation. Experimental results demonstrate a significant improvement in prediction accuracy, with average error reduced from 9.94% to 6.3%. This enhanced model offers a more efficient and accurate solution for aligning educational outcomes with industry needs.

Amnesty international is recognized as a key force in promoting social development, with higher education also facing the need for innovation. This paper explores new opportunities in educational theory and policy proposed in a recent initiative. The proposal emphasizes filtering ideology, political education, and public opinion to enhance the accuracy of ideological and political teaching. By incorporating personal suggestions through interviews, the model recommends learning materials tailored to student characteristics. System implementation and testing demonstrate its potential as a core tool for ideological education in colleges, supporting the integration of knowledge, politics, and technology to meet students’ educational needs.

Networks with smaller strong diameters generally have better fault tolerance because they enable closer connections between vertices, leading to shorter information paths. This allows the network to maintain communication and functionality more effectively during attacks or failures. In contrast, larger strong diameters mean vertices are connected over longer distances, increasing vulnerability to disruptions. Thus, the strong diameter is a key metric for assessing and optimizing network fault tolerance. This paper determines the optimal orientations for the Cartesian and strong products of even cycles, provides the minimum strong diameters and their bounds under specific conditions, and establishes a lower bound for the maximum strong diameter. A conjecture about the exact value of the maximum strong diameter is also proposed.

For a graph \(F\) and a positive integer \(t\), the edge-disjoint Ramsey number \(ER_t(F)\) is the minimum positive integer \(n\) such that every red-blue coloring of the edges of the complete graph \(K_n\) of order \(n\) results in \(t\) pairwise edge-disjoint monochromatic copies of a subgraph isomorphic to \(F\). Since \(ER_1(F)\) is in fact the Ramsey number of \(F\), this concept extends the standard concept of Ramsey number. We investigate the edge-disjoint Ramsey numbers \(ER_t(K_{1, n})\) of the stars \(K_{1, n}\) of size \(n\). Formulas are established for \(ER_t(K_{1, n})\) for all positive integers \(n\) and \(t = 2, 3, 4\) and bounds are presented for \(ER_t(K_{1, n})\) for all positive integers \(n\) and \(t \ge 5\). Furthermore, exact values of \(ER_t(K_{1, n})\) are determined for \(n = 3, 4\) and several integers \(t \ge 5\).

The development of artificial intelligence enables computers to not only simulate human artistic creations, but also synthesize fine art works with deeper meanings based on natural images. This study digitally parses the fusion of fine art and philosophy visual expressions, and develops a visual expression system based on the fusion of fine art and philosophy by utilizing a variety of key big data algorithms for visual expressions such as adversarial networks. Research on pattern recognition of this system in art creation is carried out through model training, recommendation performance evaluation, pattern recognition strategy application and regression analysis. The model in this paper works best when the number of nearest neighbors k=15, and the recommendation model in this paper can provide a personalized list of artwork recommendations for different people. The recognition of the system in this paper in the five dimensions of “spiritual level”, “value level”, “philosophical level”, “aesthetic level” and “technical level” is distributed between 4.24\(\mathrm{\sim}\)4.55. The results of regression analysis indicated that the system in this paper can improve the artistic creation as well as pattern recognition.

With development of Internet of Things, big data and artificial intelligence, cell phone signaling data, point-of-interest data and machine learning methods have been widely used in research of various fields of transportation. The use of big data processing techniques and machine learning methods to mine intercity travel data collected by various types of traffic detectors provides a new way of thinking to study travel mode selection behavior. In this paper, we pre-processed cell phone signaling data, geospatial data and interest point data around three aspects: personal attributes, travel attributes and travel mode attributes, and designed intercity travel target group extraction, travel chain extraction, travel mode extraction and travel purpose extraction algorithms, which provide basis for travel feature analysis and travel mode choice behavior prediction modeling.

Conventional techniques to electric power network (EPN) design and management are insufficient to handle extreme weather events like hurricanes due to the growing complexity and fragility of power systems. As a sophisticated simulation and optimization tool, digital twin (DT) technology may offer real-time power infrastructure monitoring and prediction. This study aims to investigate the possible application of digital twin technology in enhancing power system resilience and streamlining the design process, as well as to use it for the 3D design of the full substation engineering infrastructure process. A digital twin-based EPN model that incorporates all of the main components of the power system—power plants, substations, transmission and distribution networks, and customers—is proposed in this paper. Every component of the power system undergoes vulnerability analysis, and the chance of the system failing is calculated using a Bayesian network (BN) model and a parametric vulnerability function. According to modeling projections, Hurricane Ike will cause the majority of consumers’ power supplies to be interrupted. The model predicts that power consumption for residential, commercial, and industrial buildings will be 96.4%, 96.0%, and 94.2%, respectively, depending on the kind of building.

Radar ranging and speed measurement are common applications in daily life, with performance largely dependent on the radar signal processor. However, existing civilian radar signal processors struggle with weak signal reception and low analysis efficiency. This study designs a high-speed radar signal processor based on FPGA architecture, incorporating a fusion processing algorithm to integrate different radar signal bands, enhancing processing efficiency and accuracy. The design includes data feature analysis, storage, and fusion modules. Tests showed that the processor achieved real-time performance with a processing time under 1ms, a ranging error below 1m, and speed measurement accuracy within 5m/s, meeting practical requirements.

Intriguing symmetries are uncovered regarding all magic squares of orders 3, 4, and 5, with 1, 880, and 275,305,224 distinct configurations, respectively. In analogy with the travelling salesman problem, the distributions of the total topological distances of the paths travelled by passing through all the vertices (matrix elements) only once and spanning all elements of the matrix are analyzed. Symmetries are found to characterise the distributions of the total topological distances in these instances. These results raise open questions about the symmetries found in higher-order magic squares and the formulation of their minimum and maximum total path lengths.

In this paper, we introduce the concept of vertex-edge locating Roman dominating functions in graphs. A vertex-edge locating Roman dominating \({(ve-LRD)}\) function of a graph \(G=(V,E)\) is a function \(f:V(G)\rightarrow\{0,1,2\}\) such that the following conditions are satisfied: (i) for every adjacent vertices \(u,v\) with \(f(u)=0\) or \(f(v)=0\), there exists a vertex \(w\) at distance \(1\) or \(2\) from \(u\) or \(v\) with \(f(w)=2\), (ii) for every edge \(uv\in E\), \(\max[f(u),f(v)]\neq 0\), and (iii) any pair of distinct vertices \(u,v\) with \(f(u)=f(v)=0\) does not have a common neighbour \(w\) with \(f(w)=2\). The weight of ve-LRD function is the sum of its function values over all the vertices. The vertex-edge locating Roman domination number of \(G\), denoted by \(\gamma_{veLR}(G)\), is the minimum weight of a {ve-LRD} function in \(G\). We proved that the vertex-edge locating Roman domination problem is NP-complete for bipartite graphs. Also, we present the upper and lower bounds of \({ve-LRD}\) function for trees. Lastly, we give the upper bounds of \({ve-LRD}\) function for some connected graphs.

Traditional personnel recruitment methods are often inefficient and struggle to find candidates who meet job requirements. In this paper, we first develop a comprehensive personnel management system for colleges and universities that streamlines the recruitment process and information management. Next, recruitment data from the system is analyzed using the fuzzy C-means algorithm to cluster applicant profiles and extract position-specific user characteristics. Finally, a joint embedded neural network is employed to match applicant profiles with job positions by optimizing an objective function. Experimental results demonstrate a high job matching rate (up to 98.1%), a significantly reduced recruitment cycle (from job posting to candidate onboarding in 25 days), and a system response time as low as 0.5 seconds. These findings highlight the effectiveness of big data technology in providing timely feedback, reducing recruitment costs and staff workload, and promoting the intelligent development of talent recruitment.

The rapid development of information technology makes intelligent decision support system play an increasingly important role in economic standardized management. The Intelligent Decision Support System (IDSS) constructed in this paper includes interaction layer, analysis layer and data layer. The system standardizes the management of enterprise economy through strategic forecasting and decision analysis, economic planning and control, and economic analysis. The study combines the fuzzy hierarchical analysis method (FAHP) and the fuzzy comprehensive evaluation method (FCE) to evaluate the standardized level of economic management of enterprise A. The evaluation score of the standardized level of enterprise A’s economic management is \(F=80.955\), which is greater than 80, and it belongs to the grade of “good”. It shows that the intelligent decision support system constructed based on this paper can effectively help standardize the management of enterprise economy.

In order to solve the multi-objective optimization problem of resource allocation in enterprise strategic management, the article firstly establishes a multi-objective resource allocation model for maximizing the benefits of enterprises in enterprise strategic management. Then, it optimizes and improves the initial population, convergence factor and dynamic weights of the gray wolf algorithm, increases the population diversity by using the population strategy of reverse learning, improves the convergence factor into a nonlinear factor, and finally changes the decision-making weights of the gray wolf leadership and applies the dynamic weights to improve the accuracy of the algorithm. Subsequently, the improved gray wolf algorithm is utilized for model decoupling. By applying this paper’s algorithm and the other two algorithms to solve the six algorithms 30*6, 60*6, 90*2, 90*4, 150*4 and 150*6 for 9 times, it is found that in the analysis of the 30*6 algorithm, the enterprise’s resource allocation reaches 5,000 when the time is 110 s. At the same time, this paper’s algorithm obtains a better non-dominated solution than the other two algorithms, which proves that this paper’s algorithm solves the multi-objective resource allocation problem of enterprise law industry is proved to be effective.

In food processing, foreign matter inevitably contaminates packaged food. To ensure food safety, ray-based detection is used; however, the original images suffer from aberrations and noise that degrade quality and hinder further processing. Thus, images are preprocessed to enhance quality by highlighting key features and suppressing irrelevant ones before abnormal pattern recognition. Following image segmentation, a BP neural network algorithm is applied for foreign object detection. In tests with contaminants such as metal wires, stones, and glass, the algorithm identified distinct abnormal fluctuations at gray levels of 132, 108, and 34, respectively, allowing it to reliably detect foreign objects. Although the practical detection rate reached 100%, occasional misjudgments suggest that further optimization is needed. Overall, this method introduces a novel approach to detecting foreign objects in food and offers promising new strategies for improving food safety monitoring.

In 2003, the frequency assignment problem in a cellular network motivated Even et al. to introduce a new coloring problem: Conflict-Free coloring. Inspired by this problem and by the Gardner-Bodlaender’s coloring game, in 2020, Chimelli and Dantas introduced the Conflict-Free Closed Neighborhood \(k\)-coloring game (CFCN \(k\)-coloring game). The game starts with an uncolored graph \(G\), \(k\geq 2\) different colors, and two players, Alice and Bob, who alternately color the vertices of \(G\). Both players can start the game and respect the following legal coloring rule: for every vertex \(v\), if the closed neighborhood \(N[v]\) of \(v\) is fully colored then there exists a color that was used only once in \(N[v]\). Alice wins if she ends up with a Conflict-Free Closed Neighborhood \(k\)-coloring of \(G\), otherwise, Bob wins if he prevents it from happening. In this paper, we introduce the game for open neighborhoods, the Conflict-Free Open Neighborhood \(k\)-coloring game (CFON \(k\)-coloring game), and study both games on graph classes determining the least number of colors needed for Alice to win the game.

This paper investigates the number of rooted biloopless nonseparable planar near-triangulations and presents some formulae for such maps with three parameters: the valency of root-face, the number of edges and the number of inner faces. All of them are almost summation-free.

A graph is 1-planar if it can be drawn on the plane so that each edge is crossed by at most one other edge. In this paper, we confirm the total-coloring conjecture for 1-planar graphs without 4-cycles with maximum degree \(\Delta\geq10\).

For a graph \(G=(V,E)\) of size \(q\), a bijection \(f : E \to \{1,2,\ldots,q\}\) is a local antimagic labeling if it induces a vertex labeling \(f^+ : V \to \mathbb{N}\) such that \(f^+(u) \ne f^+(v)\), where \(f^+(u)\) is the sum of all the incident edge label(s) of \(u\), for every edge \(uv \in E(G)\). In this paper, we make use of matrices of fixed sizes to construct several families of infinitely many tripartite graphs with local antimagic chromatic number 3.

An outer independent double Roman dominating function (OIDRDF) of a graph \( G \) is a function \( f:V(G)\rightarrow\{0,1,2,3\} \) satisfying the following conditions:
(i) every vertex \( v \) with \( f(v)=0 \) is adjacent to a vertex assigned 3 or at least two vertices assigned 2;
(ii) every vertex \( v \) with \( f(v)=1 \) has a neighbor assigned 2 or 3;
(iii) no two vertices assigned 0 are adjacent.
The weight of an OIDRDF is the sum of its function values over all vertices, and the outer independent double Roman domination number \( \gamma_{oidR}(G) \) is the minimum weight of an OIDRDF on \( G \). Ahangar et al. [Appl. Math. Comput. 364 (2020) 124617] established that for every tree \( T \) of order \( n \geq 4 \), \( \gamma_{oidR}(T)\leq\frac{5}{4}n \) and posed the question of whether this bound holds for all connected graphs. In this paper, we show that for a unicyclic graph \( G \) of order \( n \), \( \gamma_{oidR}(G) \leq \frac{5n+2}{4} \), and for a bicyclic graph, \( \gamma_{oidR}(G) \leq \frac{5n+4}{4} \). We further characterize the graphs attaining these bounds, providing a negative answer to the question posed by Ahangar et al.

Let \(G\) be a \((p,q)\) graph. Let \(f\) be a function from \(V(G)\) to the set \(\{1,2,\ldots, k\}\) where \(k\) is an integer \(2< k\leq \left|V(G)\right|\). For each edge \(uv\) assign the label \(r\) where \(r\) is the remainder when \(f(u)\) is divided by \(f(v)\) (or) \(f(v)\) is divided by \(f(u)\) according as \(f(u)\geq f(v)\) or \(f(v)\geq f(u)\). \(f\) is called a \(k\)-remainder cordial labeling of \(G\) if \(\left|v_{f}(i)-v_{f}(j)\right|\leq 1\), \(i,j\in \{1,\ldots , k\}\) where \(v_{f}(x)\) denote the number of vertices labeled with \(x\) and \(\left|\eta_{e}(0)-\eta_{o}(1)\right|\leq 1\) where \(\eta_{e}(0)\) and \(\eta_{o}(1)\) respectively denote the number of edges labeled with even integers and number of edges labeled with odd integers. A graph with admits a \(k\)-remainder cordial labeling is called a \(k\)-remainder cordial graph. In this paper we investigate the \(4\)-remainder cordial labeling behavior of Prism, Crossed prism graph, Web graph, Triangular snake, \(L_{n} \odot mK_{1}\), Durer graph, Dragon graph.

Given a connected graph \(H\), its first Zagreb index \(M_{1}(H)\) is equal to the sum of squares of the degrees of all vertices in \(H\). In this paper, we give a best possible lower bound on \(M_{1}(H)\) that guarantees \(H\) is \(\tau\)-path-coverable and \(\tau\)-edge-Hamiltonian, respectively. Our research supplies a continuation of the results presented by Feng et al. (2017).

The degree of an edge \(uv\) of a graph \(G\) is \(d_G(u)+d_G(v)-2.\) The degree associated edge reconstruction number of a graph \(G\) (or dern(G)) is the minimum number of degree associated edge-deleted subgraphs that uniquely determines \(G.\) Graphs whose vertices all have one of two possible degrees \(d\) and \(d+1\) are called \((d,d+1)\)-bidegreed graphs. It was proved, in a sequence of two papers [1,17], that \(dern(mK_{1,3})=4\) for \(m>1,\) \(dern(mK_{2,3})=dern(rP_3)=3\) for \(m>0, ~r>1\) and \(dern(G)=1\) or \(2\) for all other bidegreed graphs \(G\) except the \((d,d+1)\)-bidegreed graphs in which a vertex of degree \(d+1\) is adjacent to at least two vertices of degree \(d.\) In this paper, we prove that \(dern(G)= 1\) or \(2\) for this exceptional bidegreed graphs \(G.\) Thus, \(dern(G)\leq 4\) for all bidegreed graphs \(G.\)

A proper total coloring of a graph \( G \) such that there are at least 4 colors on those vertices and edges incident with a cycle of \( G \), is called an acyclic total coloring. The acyclic total chromatic number of \( G \), denoted by \( \chi^{”}_{a}(G) \), is the smallest number of colors such that \( G \) has an acyclic total coloring. In this article, we prove that for any graph \( G \) with \( \Delta(G)=\Delta \) which satisfies \( \chi^{”}(G)\leq A \) for some constant \( A \), and for any integer \( r \), \( 1\leq r \leq 2\Delta \), there exists a constant \( c>0 \) such that if \( g(G)\geq\frac{c\Delta}{r}\log\frac{\Delta^{2}}{r} \), then \( \chi^{”}_{a}(G)\leq A+r \).

An \( (n,r) \)-arc in \( \operatorname{PG}(2,q) \) is a set \( \mathcal{B} \) of points in \( \operatorname{PG}(2,q) \) such that each line in \( \operatorname{PG}(2,q) \) contains at most \( r \) elements of \( \mathcal{B} \) and such that there is at least one line containing exactly \( r \) elements of \( \mathcal{B} \). The value \( m_r(2,q) \) denotes the maximal number \( n \) of points in the projective geometry \( \operatorname{PG}(2,q) \) for which an \( (n,r) \)-arc exists. We show by systematically excluding possible automorphisms that putative \( (44,5) \)-arcs, \( (90,9) \)-arcs in \( \operatorname{PG}(2,11) \), and \( (39,4) \)-arcs in \( \operatorname{PG}(2,13) \)—in case of their existence—are rigid, i.e. they all would only admit the trivial automorphism group of order \( 1 \). In addition, putative \( (50,5) \)-arcs, \( (65,6) \)-arcs, \( (119,10) \)-arcs, \( (133,11) \)-arcs, and \( (146,12) \)-arcs in \( \operatorname{PG}(2,13) \) would be rigid or would admit a unique automorphism group (up to conjugation) of order \( 2 \).

Let \( S \) be a connected union of finitely many \( d \)-dimensional boxes in \( \mathbb{R}^d \) and let \( \mathcal{B} \) represent the family of boxes determined by facet hyperplanes for \( S \), with \( \mathcal{F} \) the associated family of faces (including members of \( \mathcal{B} \)). For set \( F \) in \( \mathcal{F} \), point \( x \) relatively interior to \( F \), and point \( y \) in \( S \), \( x \) sees \( y \) via staircase paths in \( S \) if and only if every point of \( F \) sees \( y \) via such paths. Thus the visibility set of \( x \) is a union of members of \( \mathcal{F} \), as is the staircase kernel of \( S \). A similar result holds for \( k \)-staircase paths in \( S \) and the \( k \)-staircase kernel of \( S \).

The minimum dominating set problem asks for a dominating set with minimum size. First, we determine some vertices contained in the minimum dominating set of a graph. By applying a particular scheme, we ensure that the resulting graph is 2-connected and the length of each formed induced cycle is 0 mod 3. We label every three vertices in the induced cycles of length 0 mod 3. Then there is a way of labeling in which the set of all labeled vertices is the minimum dominating set of the resulting graph, and is contained in the minimum dominating set of the original graph. We also consider the remaining vertices of the minimum dominating set of the original graph and determine all vertices contained in the minimum dominating set of a graph with maximum degree 3. The complexity of the minimum dominating set problem for cubic graphs was shown to be APX-complete in 2000 and this problem is solved by our arguments in polynomial time.

In this paper we study a new graph parameter, the stacking number. Defined in relation to the eternal domination game, we show that there are highly connected graphs for which it is beneficial to allow multiple guards to occupy a vertex, answering an open question of Finbow et al. In fact, we show that for any sequence \( (s_i) \), allowing \( s_j \) guards to occupy a vertex can save arbitrarily many guards in comparison to allowing fewer than this on a vertex. We also show that the stacking number is \( 1 \) for all trees.

The body language of dancers is vital for conveying emotion. In this study, Kinect is used to detect and track dancers’ movements, and we develop two models: a dance action recognition model based on skeleton data and a dance emotion recognition model using an Attention-ConvLSTM. The action recognition model achieves 88.34% accuracy—reaching its best performance after just 40 iterations—while the emotion recognition model reaches an accuracy of 98.95%. Our analysis shows that features such as eigenvalue speed, skeleton pair distance, and inclination effectively differentiate emotions, although certain emotions (e.g., Excited vs. Pleased and Relaxed vs. Sad) can be confused. Notably, the leg’s skeletal points significantly influence emotion expression. Ultimately, the study establishes a dance emotion expression mechanism through coordinated movement changes of the head, hands, legs, waist, and torso.

Deep learning-based target detection algorithms outperform traditional methods by eliminating the need for manual feature design and improving accuracy and efficiency. This paper constructs a YOLOv5 target detection model using a deep convolutional neural network. To enhance accuracy, generalization, and detection speed, three data augmentation techniques—mosaic data enhancement, adaptive anchor frame, and adaptive image scaling—are applied. The model is further optimized with an attention mechanism and a modified YOLOv5 framework. A loss function and global average pooling enhance feature mapping for a fully convolutional network. Experimental results show that the improved YOLOv5n model achieves a 2.9979 percentage point increase in MAP, a 31% improvement in FPS, and a training time reduction of 10 minutes, completing 100 rounds in 20 minutes.

Metacognition, as a fundamental ability for learners to adapt to complex environments, is equally adapted to constructivist teaching and learning activities. In this paper, we propose a model of learning environment characteristics for metacognitive regulation under constructivist learning theory, and utilize Item2Vec algorithm, Self-Attention mechanism, and BiGRU model to construct a model of metacognitive ability. The model presents a kind of multi-channel network characteristic composed of Self-Attention mechanism and BiGRU model. Design a theoretical model of the learning environment oriented to improving students’ metacognitive ability, and analyze the functional modules of the overall system of the learning environment. Propose a learning activity aiming at the improvement of metacognitive ability and incorporating constructivist theory as the guiding concept to allocate the various aspects of the whole constructivist teaching activity. Analyze the implementation effect of constructivist teaching activities based on metacognitive strategies and organize the influencing factors of metacognitive strategies. The bivariate correlation analysis of students’ total test scores and usual grades are closely related to planning strategies, monitoring strategies, and regulating strategies, and the significance (two-tailed) is less than 0.01. This indicates that the higher the students’ scores, the higher the corresponding level of metacognitive strategies.

Civil engineering crack detection faces challenges due to complex environments and external interferences. This paper proposes an improved YOLO v8s-WOMA network, integrating ODConv, C2f-MA modules, and WIoU loss function to enhance crack identification accuracy. A BP neural network is also trained to assess crack damage. Experiments on the CBP dataset compare this method with existing detection algorithms. Results show that the proposed model achieves the highest mAP (90.5%), F1-score (90.3%), and accuracy (89.6%). Bridge crack detection errors remain within 0.1mm (width) and 20mm (length), ensuring precise damage assessment. The model effectively handles complex backgrounds, accurately detects cracks, and meets practical engineering needs.

The rapid growth of multilingual information online has made traditional translation insufficient, highlighting the need for intelligent language translation. This study employs a convolutional neural network to extract visual features from translated images and uses region-selective attention to align text and image features. The fused information is then processed through a sequence model to develop a computer vision-based translation algorithm. Results show that the proposed algorithm excels in key evaluation metrics, improving translation quality. It maintains a low leakage rate (1.30%), a mistranslation rate of 2.64%, and an average response time of 67.28ms. With strong generalization and applicability in multilingual translation, the algorithm demonstrates high performance and promising real-world applications.

This paper addresses the limitations of the traditional portfolio theory centered on the mean-variance model and expected utility theory, and proposes the establishment of a portfolio model that takes into account the subjective psychological factors of investors, taking into account the fact that investors are susceptible to the influence of various psychological biases, affective biases, and cognitive biases in the actual decision-making process, with respect to the theory of consistency of the assumptions of the investor’s risk attitude. The portfolio model based on fuzzy decision-making is proposed, combined with the development and application of linear programming in portfolio optimization, the return of assets is regarded as a random fuzzy variable, and the stochastic fuzzy portfolio model is constructed to consider the risk characteristics of investors. The portfolio returns under different emotions or different risk preferences are explored separately. Combined with the fund categorization allocation of the sample firms, the fund portfolio C based on the fuzzy portfolio model is proposed and compared with the equal weight allocation fund (fund portfolio A) and the risk coefficient weighted allocation fund (fund portfolio B) based on the risk level of return, respectively. Fund Portfolio C has the highest average return.

Given a prime \( p \), a \( p \)-smooth integer is an integer whose prime factors are all at most \( p \). Let \( S_p \) be the multiplicative subgroup of \( \mathbb{Q} \) generated by \(-1\) and the \( p \)-smooth integers. Define the \( p \)-smooth partial field as \( \mathbb{S}_p = (\mathbb{Q}, S_p) \). Let \( g \) be the golden ratio \( (1+\sqrt{5})/2 \). Let \( G_p \) to be the multiplicative subgroup of \( \mathbb{R} \) generated by \( g \), \(-1\), and the \( p \)-smooth integers. Define the \( p \)-golden partial field as \( \mathbb{G}_p = (\mathbb{R}, G_p) \). The partial field \( \mathbb{S}_2 \) is actually the well-known dyadic partial field and \( \mathbb{S}_3 \) has sometimes been called the Gersonides partial field. We calculate the fundamental elements of \( \mathbb{S}_5 \), \( \mathbb{G}_2 \), \( \mathbb{G}_3 \), and \( \mathbb{G}_5 \).
Our proofs make use of the SageMath computational package.