In the increasingly complex field of career planning, it has become a challenge to provide accurate and personalized advice to individuals. This study proposes a career planning assistance platform that combines microservice architecture, fuzzy logic, and deep learning technology. The platform realizes the automated processing of the whole process from user information collection to intelligent recommendation through a four-layer architecture of user interface layer, business logic layer, data processing layer, and service coordination layer. In particular, we developed a fuzzy logic model to process the fuzzy information submitted by users and innovatively designed a deep learning model with multimodal fusion to capture the complex relationships among individual interests, abilities, and career trends. The experimental results show that compared with the traditional recommender system, the model shows significant improvement in metrics such as accuracy, recall, F1 score, and root-mean-square error. In addition, through case studies, we verified the effectiveness of the model in personalized recommendation and obtained positive feedback from users. The results show that the platform not only improves the intelligence of career planning, but also provides new ideas for future research and development.

In recent years, artificial intelligence has become increasingly important for understanding the real world and the innate human ability to process intuitive physics with computers. Most research on real-world detection and prediction has used methods that generate predictive images of the environment from changes in the pixels of an image, or that predict changes in objects in the environment from changes in the numerical values of a physical simulator. However, the actual method of predicting the environment in humans is believed to consist of both visual and physical information. Therefore, in this study, in order to recognize the motion of objects, the relationship of motion characteristics between objects is represented by a graph structure, which enables the extraction of anomalous points in response to changes in the environment. Then, based on both visual and physical information, we constructed a change point prediction model that can predict the motion of objects in the environment and capture the timing of collisions between objects. Moreover, to improve the accuracy, the basic prediction model was modified and the accuracy of the model was compared. The results showed that visual information and physical information were inferred independently, and the more physical information included the physical properties of the objects, the higher the accuracy. In the change point prediction model, as the prediction accuracy of the base prediction model increased, the change point extraction accuracy also increased.

In this article, the maximum likelihood estimators (MLE) and modified maximum likelihood estimators (MMLE) of the scale and location parameters for the logistic distribution applying simple random sampling (SRS) approach and different sampling schemes of ranked set sampling (RSS) are obtained. The corresponding MLE and MMLE using RSS when the ranking is imperfect are considered too.

Smallholders play a vital role in global agriculture but face challenges such as resource limitations and market uncertainties. Contract farming (CF) and cooperative markets provide opportunities to smallholders to address these challenges. In this context, this study develops mathematical planning models for fresh produce supply chains (FPSCs) in two scenarios involving CF and cooperative markets with the objective of maximizing profits of smallholders. Results of computational experiments and sensitivity analysis reveal that CF has greater potential than cooperative markets to mitigate price volatility, thereby offering a stable income source to smallholders. This study provides valuable insights into CF and cooperative market practices, contributing to the sustainability of FPSCs and improving smallholders’ livelihoods.

With the development of online education technology, the status of English online education system is also increasing. However, the current intelligent learning guide design lacks self-adaptation and cannot get the learning state of different students. In addition, it is not possible to provide personalized learning and topic push for students with different learning states. Therefore, we propose an intelligent guide design for English online education based on deep learning, which combines knowledge tracking algorithm and exercise recommendation algorithm. Our model can extract students’ knowledge state and ability level and recommend appropriate exercises. In addition, we also introduce knowledge graph technology and use knowledge graph embedding technology to preprocess the knowledge points of exercises to enhance the state representation in the model and explore the implicit relationship. Experiments are designed to prove the effectiveness and superiority of our proposed method. Experimental results of DACK model: ACC = 93.1%, AUC = 98.8%, significantly superior to traditional knowledge tracking methods. Experimental results of DRSS model: Hit-Ratio@10 = 41.6%, NDCG@10 = 70.2%, performed excellently in personalized practice recommendations.

This study investigates user experience on the Japanese C2C platform, Mercari, through text mining analysis of App Store reviews. Using Python for data collection and KH Coder for statistical text analysis, 2,976 valid reviews were examined to uncover factors affecting user satisfaction. Keywords, co-occurrence networks, and correspondence analysis identified key areas such as item listing, purchasing processes, fees, customer service, and platform management. Positive ratings highlighted smooth transactions, engaging campaigns, and reliable service as strengths that maintained user interest. In contrast, negative reviews cited issues like inadequate customer support, high fees, and app functionality problems, leading to dissatisfaction and reduced trust. The co-occurrence network analysis depicted user concerns, particularly about fees, security, and transaction transparency. Correspondence analysis showed emotional responses driving ratings: lower scores linked to poor customer support and technical issues, while higher ratings were connected to promotions and positive shopping experiences. Cross-tabulation analysis revealed significant differences in user satisfaction among categories, emphasizing the need for operational improvements, especially in customer service and fee structures. The results indicate that enhancing customer support, refining fee policies, and ensuring smooth app usability are essential to building trust and improving satisfaction. These insights provide valuable direction for C2C platforms seeking to optimize user experience and maintain a competitive edge in the market.

Fetal growth restriction (FGR) is a disease during pregnancy that increases the risk of preterm birth and perinatal death. Currently, the diagnosis of FGR relies on ultrasonography-based estimated fetal body weight (EFBW). However, EFBW can only provide an indirect assessment of FGR because a low EFBW is only a result of growth restriction. Recent research has indicated that placental oxygenation function is a key indicator of fetal growth; however, its assessment through ultrasonography is impractical. Techniques other than ultrasonography for placental function have been investigated, and a significant difference in placental oxygenation function between FGR and non-FGR cases has been demonstrated using blood oxygen level-dependent magnetic resonance imaging (BOLD MRI). BOLD MRI can visualize oxygenation in vivo, and may be useful as a marker for the direct assessment of placental oxygenation function. However, visual assessment of placental oxygenation in BOLD MRI is challenging, even for experts, because of the complexity of analyzing the signal intensity on MRI. In this study, we proposed an automated method for predicting FGR by utilizing placental image features extracted from BOLD MRI during oxygen administration. In addition to the FGR/non-FGR classification method, we propose a placental region segmentation method to reduce the manual annotation burden. The proposed segmentation method achieved a Dice coefficient of 0.809, outperforming other deep learning methods. In the FGR/non-FGR classification experiments, conducted with four-fold cross-validation on 22 subjects, the highest performance was obtained using a pre-trained ResNet50 combined with a fully connected layer with transfer learning as a feature extractor (subject-wise accuracy of 0.908, ROC-AUC of 0.927, and F1 score of 0.922). These results demonstrate that placental image features extracted from BOLD MRI can effectively differentiate between FGR and non-FGR cases, suggesting the potential for a direct and automated approach to assess FGR through placental oxygenation function.

Deploying monocular 3D object detection networks on visual sensors of intelligent transportation assistance devices is a cost-effective and practical solution. Despite the progress made in existing monocular 3D object detection methods, there still exists a certain gap in the detection accuracy compared to 3D object detection methods based on point cloud data from LiDAR (light detection and ranging) sensors. Additionally, these methods incur relatively high computational costs. Addressing these issues, this paper proposes an improved monocular 3D object detection network, which optimizes the overall structure of the model through structural reparameterization, effectively alleviating the computational burden on computing devices. Simultaneously, we focus on the differences between 2D and 3D features and propose a cross-dimension focusing method to enhance the performance of ceiling the object detection method in extracting 3D object features. In the KITTI benchmarks, our framework achieved significantly superior performance in 3D object detection compared to other methods.

The intersection of deep learning (DL) techniques and electroencephalogram (EEG) signals to predict multiple human sleep behaviors is integrated in this study. In fact, the detection of multiple sleep behaviors is a critical task for mental health professionals. To address this issue, an advanced analytical framework based on DL is developed in this study. After pre-processing the signals, this study developed a novel approach on 1D-EEG signals, which are converted to the series of (3x times) stack images using ensemble of short-time Fourier transform (STFT), continuous wavelet transform (CWT), and mel-frequency cepstral coefficients (MFCCs) techniques. Later one, the improved MaxViT model is improved to extract features from 3D-cum-stacked signal images. To capture the temporal dynamics of sleep architecture, we then implemented a gated recurrent unit (GRU) model. The AdaBoost classifier was finally applied for the classification of multiclass sleep behaviors. The performance of the proposed system (mSleep) is validated by using PhysioNet dataset. Our study introduces a novel multi-modal feature extraction approach by combining STFT, CWT, and MFCCs to create 3D-stacked EEG representations, enabling comprehensive temporal-spectral analysis. We develop a hybrid DL pipeline integrating Enhanced MaxViT for feature extraction, GRU for temporal modeling, and AdaBoost for multi-class classification, achieving superior accuracy and generalization. Additionally, we address class imbalance using SMOTE and enhance model interpretability through AdaBoost, ensuring clinical applicability and robust performance on the PhysioNet Sleep-EDF dataset. The mSleep system has broad implications for sleep research and neural signal processing, as it provides information that can be useful in the design of individualized therapy plans or technological improvement plans. The code and dataset can be downloaded from https://www.github.com/qaisar256/sleepdisorder1.0/.

Stereo-matching has become essential in various industrial applications, including robotics, autonomous driving, and drone-based surveying. In the drone-based depth estimation, we captured images from two different positions and determined the corresponding points between them through stereo-matching. A longer distance between the two positions improves triangulation accuracy but makes stereo-matching difficult owing to the reduced image overlap. This limitation is inherent to previous methods, necessitating at least 50% image overlap to achieve only centimeter-level accuracy. Hence, we propose using stereo viewing with feature point matching, which allows for direct matching of points on the image. Our approach applies a novel rotation-invariant convolutional neural network (CNN) that extracts features more effectively in the presence of angular changes in a subject, surpassing the performance of previous CNN-based models. We evaluated our method using the HPatches dataset, which demonstrated an increase in feature point matching accuracy of up to 0.9%. In a practical stereo imaging setting, our method achieved a height estimation error of approximately 1.2 mm and height resolution of approximately 2.6 mm in image pairs with approximately 25% overlap under varying conditions. This performance confirms that the proposed approach effectively resolves the trade-off inherent to traditional stereo-matching techniques, particularly with regard to the challenging overlapping scenarios that these previous methods failed to account for. Consequently, this study substantially broadens the applicability and versatility of stereo-depth estimation.

According to Japan’s Ministry of Health, Labour and Welfare, work-related lower back pain is prevalent in industries such as commerce, finance, health, and hygiene. Such pain is primarily caused by performing tasks involving heavy lifting and carrying in various fields, including caregiving, transportation, and agriculture. This study proposes a lumbar-powered exoskeleton to assist lifting movements. Passive exoskeletons use springs or rubber belts for assistive force, thus rendering them lightweight but unable to provide controlled assistance based on the wearer’s movements. Active exoskeletons, such as the Hybrid Assistive Limb robot suit, use surface electromyography (sEMG) to detect movement characteristics. However, sEMG is susceptible to noise owing to factors such as sweating and skin contamination. This study proposes a lumbar exoskeleton control method using a nine-axis inertial measurement unit (IMU) that is easy to attach and less affected by the wearer’s state. Convolutional neural networks and long short-term memory models are adopted for posture classification. Tests involving 10 subjects show that integrated electromyographic activity decreased significantly.

In recent years, “Tame-map” has emerged as a social media platform for local revitalization. It is a web application for sharing local information. It enables users to conveniently post and view information on local events in their daily lives. Because many “Tame-map” users are likely to participate in events, increasing the number of views is an important issue from the perspective of regional revitalization. The design relies on the organizer’s experience and intuition, and there is no established method for developing a design that attracts a large number of viewers. Therefore, if the number of visitors can be predicted in advance, it is feasible to reconsider the design of flyers based on this information. In addition, in click-through rate (CTR) prediction, which is an aspect of advertising analysis, it has been revealed that predicting the user attributes of viewers contributes to improving the prediction accuracy. However, in the “Tame-map” system, user attributes of viewers do not exist. In this study, we aim to clarify the extent to which considering the generational tags assigned to events would impact the prediction of click counts.

The fluxgate sensor is the most widely used sensor in vector magnetic measurement. However, during long-term continuous observation, the fluxgate sensor will produce large measurement errors due to changes in ambient temperature. This paper proposes a temperature calibration method for the fluxgate sensor based on polynomial fitting to address the temperature error of fluxgate sensors. First, the effect of temperature on the performance indicators of the fluxgate sensor was analyzed. Second, according to the existing temperature-magnetic field data, a temperature calibration model of the fluxgate sensor was constructed. Compared with other temperature calibration methods, the result shows that the proposed temperature calibration method is relatively simple and can better achieve real-time calibration for sensor application scenarios. Finally, to verify the effectiveness of the proposed method, numerous laboratory experiments were implemented. The temperature drift was reduced from about 700 nT before calibration to about 1 nT, root mean square error RMSE = 11.7189, indicating that the proposed method has a good temperature calibration effect on the data measured by the fluxgate sensor within the variable temperature background.

The intermittency, uncontrollability, and variability of wind power affect the economical operation and reliable delivery of the power system. To ensure the smooth integration of wind power into the grid, an accurate wind power forecast is essential. In this paper, we propose a short-term wind power prediction model based on decomposition and reconstruction to improve the accuracy of wind power forecasts. This model has the following characteristics: (i) wind power time series are decomposition into multiple components through the application of a decomposition method that combines fully integrated empirical mode decomposition with adaptive noise algorithm; (ii) the components are clustered using K-means clustering based on dynamic time warping. According to the similarity of the complexity and lag of components, they are divided into three classes; (iii) three different prediction models, including seasonal autoregressive integrated moving average model (SARIMA), long short-term memory network (LSTM) and Bi-directional long short-term memory network (BiLSTM), predict three types of components respectively. Finally, to illustrate the capability of the model, we compare its performance with three typical models. The results demonstrate that the proposed method exceeds the baseline models in regard to prediction performance.

The increasing complexity of medical tasks has imposed significant challenges on existing task allocation systems in healthcare. This study introduces a flexible multi-objective optimization model utilizing goal-directed optimization to address this challenge. The model prioritizes urgent medical tasks, aiming to minimize the potential loss of task value and reduce patient health risks. Simultaneously, it strives to optimize the use of available resources to ensure the sustainability of operations. To address the optimization problem, an Adaptive Multi-Objective Ant Colony Optimization algorithm was introduced. This approach incorporates a dynamic heuristic function and a flexible pheromone initialization mechanism to enhance the efficiency and accuracy of task allocation. The experimental results demonstrate that the proposed algorithm outperforms others in terms of convergence speed, solution quality, and flexibility, providing an effective tool for reducing the burden on medical staff and improving the operational efficiency of healthcare institutions.

As natural gas pipeline networks expand, the complexity of pipeline scheduling models increases, making feasibility analysis increasingly difficult. This study focuses on the feasibility analysis of optimization models for natural gas distribution network scheduling, considering it as a classification problem. Models grounded in traditional neural networks, parallel branch neural networks, and graph neural networks are developed and assessed. Two distinct scales of natural gas distribution networks are explored by collecting a limited dataset of sample cases to train and validate the proposed feasibility analysis models through empirical case studies. The results demonstrate that the parallel branch neural network exhibits superior predictive performance. In addition, this study introduces an innovative methodology for traceability diagnosis of infeasible cases, offering a practical framework for engineering applications.

This study analyzed the modeling and characteristics of parallel-plate capacitors for underwater capacitive wireless power transfer (CWPT) systems. Using finite element method (FEM) simulations, this study investigated the impact of various factors on the simulation results for both capacitance and conductance. A fitting equation is proposed for the coupling capacitance and conductance in a seawater environment. The derived equations were verified by varying the coupler size. For four different sizes, the maximum error percentage for capacitance was 9.18% at a size of 180×100 mm. For conductance, all error percentages were less than 3.49% at a transfer distance of 20 mm. The agreement between the simulated results and those calculated from the derived equations confirms the validity of the derived equations for both capacitance and conductance. Notably, this study also demonstrates a ratio of approximately 295.9 between the real and imaginary parts of the coupling admittance at a frequency of 3 MHz. This finding confirms that conductance, rather than susceptance, dominates the CWPT systems in underwater applications.

Ultracompact electric vehicles have compact, lightweight bodies with low outer-plate rigidity. This results in the transmission of road noise from the tires and wind noise (caused by the projection shape of the vehicle) into the cabin. Interior acoustic control systems require devices that can produce sound waves. Next-generation mobility uses giant magnetostrictive actuators (GMAs) for sound production. This is a foundational study for designing GMAs that can be used in such interior acoustic control systems. Magnetostriction forces are generated when a magnetic field deforms giant magnetostrictive materials. The output characteristics were analyzed by electromagnetic-field analysis. Further, the GMA magnetostriction force was analyzed using the finite element method (FEM)—across varying frequency ranges of piezoelectrically controlled amplifiers. The FEM results indicated that the GMA output performance was sufficient to generate sound waves for active noise control in the low-frequency range, 100–500 Hz (road noise). Further optimization is required to expand the frequency range—to accommodate music playback, etc.—including modification of the actuator size, weight, shape, and components and using materials with higher magnetic permeability.

Recently, robotic systems that offer practical services in everyday life, such as smart home systems, have been developed. The latest trend moves beyond user-controlled internal systems through traditional methods, such as remote controls, to systems that autonomously understand user contexts and evolve to provide a more comfortable living environment. Therefore, this research aims to advance this field by exploring the potential of a system capable of understanding human conditions and behavior, and by proposing or executing actions that align with an individual’s intended actions. To investigate a system capable of achieving this goal, we focused on analyzing the gazes of users and developed an eyeglass-type wearable device. The primary objective of this study was to track a specific user’s gaze, identify the object of focus, and analyze the user’s level of attention and interest in that object. Therefore, for the sensory configuration of the system, an analysis was performed using data collected from camera sensors for eye tracking and sensors for measuring environmental information. Based on the analysis results, we evaluated whether the system could accurately interpret and anticipate the actions that user intended to perform.

Suicide is more prevalent among individuals with psychiatric disorders, underscoring the importance of early identification of warning signs for intervention. Common suicide detection models for text analysis often require tremendous labeled data, making them prone to overfitting when dealing with tiny datasets. Aiming at the problem, we propose a prompt-based learning suicide detection model that is suitable in low-resource settings following the “pre-train, prompt, predict” paradigm, named E3.0-HP-SDM (ERNIE 3.0 Hybrid Prompt-Suicide Detection Model). In the construction of the E3.0-HP-SDM, we selected ERNIE 3.0, renowned for its knowledge enhancement capabilities, as our pre-trained language model (PLM). Additionally, we developed a hybrid prompt template, which integrates a set of tunable soft prompts into a specific suicide-related hard prompt template. This template reformulates the original input into a format with unfilled slots, specifically designed to guide the PLM in applying its knowledge-masked language model for the inference of suicide intentions. When tested on identical data, E3.0-HP-SDM outperforms not only other models within the same paradigm but also often-cited baseline combination models that follow the third paradigm of natural language processing, the “pre-train, fine-tune” paradigm, with an accuracy of 87.6% and an AUC of 85.2%.

Video anomaly detection is crucial in intelligent surveillance, yet the scarcity and diversity of abnormal events pose significant challenges for supervised methods. This paper presents an unsupervised framework that integrates graph attention networks (GATs) and Transformer architectures, combining masked autoencoders (MAEs) with self-distillation training. GATs are utilized to model spatial and inter-frame relationships, while Transformers capture long-range temporal dependencies, overcoming the limitations of traditional MAE and self-distillation approaches. The model employs a two-stage training process: first, a lightweight MAE combined with a GAT-Transformer fusion constructs a knowledge distillation module; second, the student autoencoder is optimized by integrating a graph convolutional autoencoder and a classification head to identify synthetic anomalies. We evaluate the proposed method on three representative datasets—ShanghaiTech Campus, UBnormal, and UCSD Ped2—and achieve promising results.

This study proposes a graph convolutional network (GCN)-based data–model interactive remaining useful life (RUL) prediction method for tools. First, a composite health indicator (CHI) is built by aggregating information from neighboring nodes through the GCN. Second, a stochastic degradation model is established to capture the time-varying evolutionary trend. Specifically, the drift coefficient is treated as a random variable to represent its variability among different individuals of the same type of tool, and the model parameters are estimated using intermediate evolutionary process data. Then, a data–model interactive mechanism is proposed by forming closed-loop optimization between the CHI construction and the stochastic degradation model to enhance the RUL prediction accuracy. Finally, experiments are conducted on the PHM2010 dataset to verify the effectiveness and superiority of the proposed method.

Digital technology innovation has become a powerful driving force for economic growth and social development. However, due to significant disparities in digital technology research and development investment, patent applications, industrial foundation, and policy support across different regions, there exists a notable variation in the spatial distribution and development level of digital economy innovation among China’s regions. This article measures the digital economy innovation levels of each region, establishes a spatial association network using the gravity model to calculate the centrality indicators of each region, and thereby determines their position in the network. The digital economy innovation levels and network positions of each region are used as core explanatory variables, and a spatial econometric model is employed to analyze their impact on economic growth. The main conclusions of the article are as follows. (1) From 2011 to 2022, the cities with the highest levels of digital economy innovation in China are concentrated in economically developed regions such as Beijing–Tianjin–Tangshan, the Yangtze River Delta, and the Pearl River Delta. The overall distribution pattern shows higher levels in the southeast and lower levels in the northwest. (2) Provinces with higher centrality in the digital economy innovation network include both economically developed and underdeveloped regions, indicating the spatial effects of digital economy innovation. (3) Digital economy innovation can not only promote local economic growth but also drive economic growth in neighboring regions through spillover effects. (4) There is a positive correlation between the position in the digital economy innovation network and regional economic growth rates.

The advent of digital economy has had a significant impact on the scale and quality of employment. This study employs a regional characteristic analysis to investigate the impact of DE development on employment quantity and quality. Additionally, the effect of DE advancement on employment levels is examined, with emphasis on the associated changes in quality. The results demonstrate that the progression of the DE exerts a dual influence on the overall employment landscape, encompassing both a “creation effect” and a “substitution effect.” However, the “creation effect” is identified as the predominant influence. The advancement of the DE has the potential to markedly enhance the quality of regional employment opportunities. The impact of digital economic development on regional employment income and security is contingent upon the level of development. The DE has a beneficial effect on the stability of employment in the eastern and central regions. Nevertheless, in the western region, the impact is less significant, with the level of employment security exhibiting a lagging phenomenon. Therefore, this study proposes a series of policy recommendations, including the acceleration of digital economic development, the expansion of new employment opportunities, enhancement of digital skills among the labor force, and the improvement of the social welfare system for workers.