Construction Of Model Research Articles

Graph neural network based human detection in videos during occlusion environments

One of the most difficult perceptual problems for many applications is accurately recognizing the human object in a variety of circumstances. This can be difficult due to obstructions, weather, complex backdrops, cast shadows, and occlusions. Occlusion is a challenging open problem where a detector can only perceive a portion of the target human because of obstacles in the surrounding. In this research, an experimental investigation was conducted using the multi object tracking (MOT17) datasets to construct a graph neural network-based solution for the detection of humans in videos while considering the possibility of occlusion. Graph neural network (GNN) is used for the construction of neural solver model for detecting human object in occlusion scenario. The results obtained shows that this proposed method offers a considerable improvement in efficiency in comparison to the ways that have been used in the past. The values obtained for the standard performance metrics are higher than the state-of-the-art methods.

Uncovering botnets in IoT sensor networks: a hybrid self-organizing maps approach

The integration of the internet of things (IoT) has revolutionized diverse industries, introducing interconnected devices and IoT sensor networks for improved data acquisition. However, this connectivity exposes IoT ecosystems to emerging threats, with botnets posing significant risks to security. This research aims to develop an innovative solution for detecting botnets in IoT sensor networks. Leveraging insights from existing research, the study focuses on designing a hybrid self-organization map (SOM) Approach that integrates lightweight deep learning (DL) techniques. The objective is to enhance detection accuracy by exploring various DL architectures. Proposed methodology aims to balance computational efficiency for resource-constrained IoT devices while improving the discriminatory power of the detection system. The study advancing IoT cybersecurity and addresses critical challenges in botnet detection within IoT sensor networks. The testing of the artificial neural networks (ANN) classifier involves three models, each represented based on parameters related to the construction of the training models. The most effective ANN achieves 86%, works on anomaly intrusion detection systems (AIDS).

Advancements and characteristics of scholar profiling research in China

This study analyzes the development and characteristics of scholar profiling research in China from 2016 to 2022, utilizing content analysis of 38 core journal articles. Scholar profiling, adapted from user profiling in big data, involves extracting and analyzing academic data to build models that describe scholars' characteristics and behaviors. Our findings highlight the use of diverse data sources, including commercial databases and social platforms, and emphasize the importance of data preprocessing and modeling techniques. While applied research dominates, the field faces challenges such as the lack of standardized data collection and depth in model construction. The study underscores the need for robust profiling tools and broader application scenarios to enhance the effectiveness of scholar profiling systems.

Clinical Application of Artificial Intelligence in the Ultrasound Classification of Hepatic Cystic Echinococcosis.

Hepatic cystic echinococcosis (HCE) is a zoonotic disease that occurs when the larvae of Echinococcus granulosus parasitize the livers of humans and mammals. HCE has five subtypes, and accurate subtype classification is critical for choosing a treatment strategy. To evaluate the clinical utility of artificial intelligence (AI) based on convolutional neural networks (CNNs) in the classification of HCE subtypes via ultrasound imaging, we collected ultrasound images from 4,012 HCE patients at the First Affiliated Hospital of Xinjiang Medical University between 2008 and 2020. Specifically, 1,820 HCE images from 967 patients were used as the training and validation sets for the construction of the AI model, and the remaining 6,808 images from 3,045 patients were used as the test set to evaluate the performance of the AI models. The 6,808 images were randomly divided into six groups, and each group contained equal proportions of the five subtypes. The data of each group were analyzed by a resident physician. The accuracy of HCE subtype classification by the AI model and by manual inspection was compared. The AI HCE classification model showed good performance in the diagnosis of subtypes CE1, CE2, CE4, and CE5. The overall accuracy of the AI classification (90.4%) was significantly greater than that of manual classification by physicians (86.1%; P <0.05). The CNN can better identify the five subtypes of HCE on ultrasound images and should help doctors with little experience in more accurately diagnosing HCE.

Integrated transcriptomic and immunological profiling reveals new diagnostic and prognostic models for cutaneous melanoma

The mortality rate associated with cutaneous melanoma (SKCM) remains alarmingly high, highlighting the urgent need for a deeper understanding of its molecular underpinnings. In our study, we leveraged bulk transcriptome sequencing data from the SKCM cohort available in public databases such as TCGA and GEO. We utilized distinct datasets for training and validation purposes and also incorporated mutation and clinical data from TCGA, along with single-cell sequencing data from GEO. Through dimensionality reduction, we annotated cell subtypes within the single-cell data and analyzed the expression of tumor-related pathways across these subtypes. We identified differentially expressed genes (DEGs) in the training set, which were further refined using the Least Absolute Shrinkage and Selection Operator (LASSO) machine learning algorithm, employing tenfold cross-validation. This enabled the construction of a prognostic model, whose diagnostic efficacy we subsequently validated. We conducted Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses on the DEGs, and performed immunological profiling on two risk groups to elucidate the relationship between model genes and the immune responses relevant to SKCM diagnosis, treatment, and prognosis. We also knocked down the GMR6 expression level in the melanoma cells and verified its effect on cancer through multiple experiments. The results indicate that the GMR6 gene plays a role in promoting the proliferation, invasion, and migration of cancer cells in human melanoma. Our findings offer novel insights and a theoretical framework that could enhance prognosis, treatment, and drug development strategies for SKCM, potentially leading to more precise therapeutic interventions.

Learning Electronic Polarizations in Aqueous Systems.

The polarization of periodically repeating systems is a discontinuous function of the atomic positions, a fact which seems at first to stymie attempts at their statistical learning. Two approaches to build models for bulk polarizations are compared: one in which a simple point charge model is used to preprocess the raw polarization to give a learning target that is a smooth function of atomic positions and the total polarization is learned as a sum of atom-centered dipoles and one in which instead the average position of Wannier centers around atoms is predicted. For a range of bulk aqueous systems, both of these methods perform perform comparatively well, with the former being slightly better but often requiring an extra effort to find a suitable point charge model. As a challenging test, we also analyze the performance of the models at the air-water interface. In this case, while the Wannier center approach delivers accurate predictions without further modifications, the preprocessing method requires augmentation with information from isolated water molecules to reach similar accuracy. Finally, we present a simple protocol to preprocess the polarizations in a data-driven way using a small number of derivatives calculated at a much lower level of theory, thus overcoming the need to find point charge models without appreciably increasing the computation cost. We believe that the training strategies presented here help the construction of accurate polarization models required for the study of the dielectric properties of realistic complex bulk systems and interfaces with ab initio accuracy.

A study on construction of a prognosis model for liver cancer based on analgesic targets and screening therapeutic drugs.

Liver cancer is one of the most malignant liver diseases in the world, and the 5-year survival rate of such patients is low. Analgesics are often used to cure pain prevalent in liver cancer. The expression changes and clinical significance of the analgesic targets (ATs) in liver cancer have not been deeply understood. The purpose of this study is to clarify the expression pattern of ATs gene in liver cancer and its clinical significance. Through the comprehensive analysis of transcriptome data and clinical parameters, the prognosis model related to ATs gene is established, and the drug information sensitive to ATs is mined. The study primarily utilized transcriptomic data and clinical information from liver cancer patients sourced from The Cancer Genome Atlas (TCGA) database. These data were employed to analyze the expression of ATs, conduct survival analysis, gene set variation analysis (GSVA), immune cell infiltration analysis, establish a prognostic model, and perform other bioinformatic analyses. Additionally, data from liver cancer patients in the International Cancer Genome Consortium (ICGC) were utilized to validate the accuracy of the model. Furthermore, the impact of analgesics on key genes in the prognostic model was assessed using data from the Comparative Toxicogenomics Database (CTD). The study investigated the differential expression of 58 ATs genes in liver cancer compared to normal tissues. Patients were stratified based on ATs expression, revealing varied survival outcomes. Functional enrichment analysis highlighted distinctions in spindle organization, centrosome, and spindle microtubule functions. Prognostic modeling identified low TP53 expression as protective, while elevated CCNA2, NEU1, and HTR2C levels posed risks. Commonly used analgesics, including acetaminophen and others, were found to influence the expression of these genes. These findings provide insights into potential therapeutic strategies for liver cancer and shed light on the molecular mechanisms underlying its progression. The collective analysis of gene signatures associated with ATs suggests their potential as prognostic predictors in hepatocellular carcinoma patients. These findings not only offer insights into cancer therapy but also provide novel avenues for the development of indications for analgesics.

Construction of multi-component finite element model to predict biomechanical behaviour of breasts during running and quantification of the stiffness impact of internal structure.

This study aims to investigate the biomechanical behaviour and the stiffness impact of the breast internal components during running. To achieve this, a novel nonlinear multi-component dynamic finite element method (FEM) has been established, which uses experimental data obtained via 4D scanning technology and a motion capture system. The data are used to construct a geometric model that comprises the rigid body, layers of soft tissues, skin, pectoralis major muscle, fat, ligaments and glandular tissues. The traditional point-to-point method has a relative mean absolute error of less than 7.92% while the latest surface-to-surface method has an average Euclidean distance (d) of 7.05mm, validating the simulated results. After simulating the motion of the different components of the breasts, the displacement analysis confirms that when the motion reaches the moment of largest displacement, the displacement of the breast components is proportional to their distance from the chest wall. A biomechanical analysis indicates that the stress sustained by the breast components in ascending order is the glandular tissues, pectoralis major muscle, adipose tissues, and ligaments. The ligaments provide the primary support during motion, followed by the pectoralis major muscle. In addition, specific stress points of the breast components are identified. The stiffness impact experiment indicates that compared with ligaments, the change of glandular tissue stiffness had a slightly more obvious effect on the breast surface. The findings serve as a valuable reference for the medical field and sports bra industry to enhance breast protection during motion.

Construction and Validation of Surface Soil Moisture Inversion Model Based on Remote Sensing and Neural Network

Surface soil moisture (SSM) reflects the dry and wet states of soil. Microwave remote sensing technology can accurately obtain regional SSM in real time and effectively improve the level of agricultural drought monitoring, and it is of great significance for agricultural precision irrigation and smart agriculture construction. Based on Sentinel-1, Sentinel-2, and Landsat-8 images, the effect of vegetation was removed by the water cloud model (WCM), and SSM was retrieved and validated by a radial basis function (RBF) neural network model in bare soil and vegetated areas, respectively. The normalized difference vegetation index (NDVI) calculated by Landsat-8 (NDVI_Landsat-8) had a better effect on removing the influence the of vegetation layer than that of NDVI_Sentinel-2. The RBF network model, established in a bare area (R = 0.796; RMSE = 0.029 cm3/cm3), and the RBF neural network model, established in vegetated areas (R = 0.855; RMSE = 0.024 cm3/cm3), have better simulation effects on SSM than a linear SSM inversion model with single polarization. The introduction of surface parameters to the RBF neural network model can improve the accuracy of the model and realize the high-accuracy inversion of SSM in the study area.

The statistical and dynamic modeling of the first part of the 2013-2014 Euromaidan protests in Ukraine: The Revolution of Dignity and preceding times.

Ukraine's tug-of-war between Russia and the West has had significant and lasting consequences for the country. In 2013, Viktor Yanukovych, the Ukrainian president aligned with Russia, opted against signing an association agreement with the European Union. This agreement aimed to facilitate trade and travel between the EU and Ukraine. This decision sparked widespread protests that coalesced in Kyiv's Maidan Square, eventually becoming known as the Euromaidan protests. In this study, we analyze the protest data from 2013, sourced from Ukraine's Center for Social and Labor Research. Despite the dataset's limitations and occasional inconsistencies, we demonstrate the extraction of valuable insights and the construction of a descriptive model from such data. Our investigation reveals a pre-existing state of self-excitation within the system even before the onset of the Euromaidan protests. This self-excitation intensified during the Euromaidan protests. A statistical analysis indicates that the government's utilization of force correlates with increased future protests, exacerbating rather than quelling the protest movement. Furthermore, we introduce the implementation of Hawkes process models to comprehend the spatiotemporal dynamics of the protest activity. Our findings highlight that, while protest activities spread across the entire country, the driving force behind the dynamics of these protests was the level of activity in Kyiv. Furthermore, in contrast to prior research that emphasized geographical proximity as a key predictor of event propagation, our study illustrates that the political alignment among oblasts, which are the distinct municipalities comprising Ukraine, had a more profound impact than mere geographic distance. This underscores the significance of social and cultural factors in molding the trajectory of political movements.

Mathematical Models of the Arabidopsis Circadian Oscillator

We review the construction and evolution of mathematical models of the Arabidopsis circadian clock, structuring the discussion into two distinct historical phases of modeling strategies: extension and reduction. The extension phase explores the bottom-up assembly of regulatory networks, introducing as many components and interactions as possible to capture the oscillatory nature of the clock. The reduction phase deals with functional decomposition, distilling complex models to their essential dynamical repertoire. Current challenges in this field, including the integration of spatial considerations and environmental influences like light and temperature, are also discussed. The review emphasizes the ongoing need for models that balance molecular detail with practical simplicity.

Detecting Road Intersections from Crowdsourced Trajectory Data Based on Improved YOLOv5 Model

In recent years, the rapid development of autonomous driving and intelligent driver assistance has brought about urgent demands on high-precision road maps. However, traditional road map production methods mainly rely on professional survey technologies, such as remote sensing and mobile mapping, which suffer from high costs, object occlusions, and long updating cycles. In the era of ubiquitous mapping, crowdsourced trajectory data offer a new and low-cost data resource for the production and updating of high-precision road maps. Meanwhile, as key nodes in the transportation network, maintaining the currency and integrity of road intersection data is the primary task in enhancing map updates. In this paper, we propose a novel approach for detecting road intersections based on crowdsourced trajectory data by introducing an attention mechanism and modifying the loss function in the YOLOv5 model. The proposed method encompasses two key steps of training data preparation and improved YOLOv5s model construction. Multi-scale training processing is first adopted to prepare a rich and diverse sample dataset, including various kinds and different sizes of road intersections. Particularly to enhance the model’s detection performance, we inserted convolutional attention mechanism modules into the original YOLOv5 and integrated other alternative confidence loss functions and localization loss functions. The experimental results demonstrate that the improved YOLOv5 model achieves detection accuracy, precision, and recall rates as high as 97.46%, 99.57%, and 97.87%, respectively, outperforming other object detection models.

Construction and validation of nomogram prognostic model for predicting survival in hepatoblastoma patients: a population-based study.

For patients with hepatoblastoma (HB), current staging system is not accurate in predicting survival outcomes. The aim of this study was to develop two accurate survival prediction models to guide clinical decision making. A retrospective analysis of 424 HB patients was performed from 2004 to 2015 using the Surveillance, Epidemiology and End Results (SEER) database. Univariate and multivariate Cox regression analysis was used to screen for variables. The identified variables were used to build survival prediction model. The performance of the nomogram models was assessed based on the concordance index (C-index), calibration plot, and receiver operating characteristic (ROC) curve. The Cox regression analysis identified six variables affecting overall survival (OS) in HB patients, including race, tumor size, lymph node involvement, distant metastases, surgery and chemotherapy. And the Cox regression analysis identified five variables including race, lymph node involvement, distant metastases, surgery, and chemotherapy that affect cancer-specific survival (CCS) in HB patients. In the training cohort, the C-index of the nomogram in predicting the OS was 0.791 [95% confidence intervals (95% CI) 0.717-0.865], CSS was 0.805(95% CI 0.728-0.882). In the validation cohort, the C-index of the nomogram in predicting the OS was 0.712 (95% CI 0.511-0.913), the CSS was 0.751 (95% CI 0.566-0.936). In the training cohort, the area under the receiver operator characteristics curve (AUC) values of the nomogram in prediction of the 1-, 3-, and 5-year OS were 0.842 (95% CI 0.739-0.944), 0.759 (95% CI 0.670-0.849), and 0.770 (95% CI 0.686-0.852), respectively. In the validation cohort, the AUC values for prediction of the 1-, 3-, and 5-year OS were 0.920 (95% CI 0.806-1.034), 0.863 (95% CI 0.750-0.976), and 0.844 (95% CI 0.721-0.967), respectively. Two nomogram models were developed and validated in this study which provided accurate prediction of the OS and CSS in HB patients. The constructed models can be used for predicting survival outcomes and guide treatment for HB patients.

MPHGCL-DDI: Meta-Path-Based Heterogeneous Graph Contrastive Learning for Drug-Drug Interaction Prediction

The combinatorial therapy with multiple drugs may lead to unexpected drug-drug interactions (DDIs) and result in adverse reactions to patients. Predicting DDI events can mitigate the potential risks of combinatorial therapy and enhance drug safety. In recent years, deep models based on heterogeneous graph representation learning have attracted widespread interest in DDI event prediction and have yielded satisfactory results, but there is still room for improvement in prediction performance. In this study, we proposed a meta-path-based heterogeneous graph contrastive learning model, MPHGCL-DDI, for DDI event prediction. The model constructs two contrastive views based on meta-paths: an average graph view and an augmented graph view. The former represents that there are connections between drugs, while the latter reveals how the drugs connect with each other. We defined three levels of data augmentation schemes in the augmented graph view and adopted a combination of three losses in the model training phase: multi-relation prediction loss, unsupervised contrastive loss and supervised contrastive loss. Furthermore, the model incorporates indirect drug information, protein–protein interactions (PPIs), to reveal latent relations of drugs. We evaluated MPHGCL-DDI on three different tasks of two datasets. Experimental results demonstrate that MPHGCL-DDI surpasses several state-of-the-art methods in performance.

Development of a biomarker signature associated with anoikis to predict prognosis and immunotherapy response in melanoma.

Skin cutaneous melanoma (SKCM) is malignant cancer known for its high aggressiveness and unfavorable prognosis, particularly in advanced tumors. Anoikis is a specific pattern of programmed cell death associated with tumor regeneration, migration, and metastasis. Nevertheless, limited research has been conducted to investigate the function of anoikis in SKCM. Anoikis-related genes (ARGs) were extracted from Genecards to identify SKCM subtypes and to explore the immune microenvironment between the different subtypes. Prognostic models of SKCM were developed by LASSO COX regression analysis. Subsequently, the predictive value of risk scores in SKCM and the association with immunotherapy were further explored. Finally, the expression of 6 ARGs involved in the model construction was detected by immunohistochemistry and PCR. This study identified 20 ARGs significantly associated with SKCM prognosis and performed disease subtype analysis of samples based on these genes, different subtypes exhibited significantly different clinical features and tumor immune microenvironment (TIME) landscapes. The risk score prognostic model was generated by further screening and identification of the six ARGs. The model exhibited a high degree of sensitivity and specificity to predict the prognosis of individuals with SKCM. These high- and low-risk populations showed different immune statuses and drug sensitivity. Further immunohistochemical and PCR experiments identified significant differential expression of the six ARGs in tumor and normal samples. Anoikis-based features may serve as novel prognostic biomarkers for SKCM and may provide important new insights for survival prediction and individualized treatment development.

Research on Promoting Entrepreneurship among Returning Migrant Workers

To encourage migrant workers to return to their hometowns for entrepreneurship, the Guangxi government has formulated a series of supportive policies under the guidance of national policies. Understanding the direction for policy improvement requires exploring what migrant workers truly want and what concerns they have. This paper first establishes an index system based on eight dimensions including subjective attitudes, physiological motivations, and social motivations through the construction of a Random Forest model, analyzing the importance of various factors affecting the willingness of migrant workers to return home for entrepreneurship. The study found that the factor "influence from peer groups or media reports" has a significant impact on their willingness. Secondly, using the AMIRA model to predict the per capita disposable income of rural residents in Guangxi for the next seven years, it was found that their income is on an upward trend, which is conducive to enhancing the willingness of migrant workers to return home for entrepreneurship. Lastly, based on the research, countermeasures and suggestions are proposed from the perspectives of the government, market, and society to promote entrepreneurship among returning migrant workers.

Construction and alidation of a severity prediction model for acute pancreatitis based on CT severity index: A retrospective case-control study.

To construct and internally and externally validate a nomogram model for predicting the severity of acute pancreatitis (AP) based on the CT severity index (CTSI).A retrospective analysis of clinical data from 200 AP patients diagnosed at the Hefei Third Clinical College of Anhui Medical University from June 2019 to June 2022 was conducted. Patients were classified into non-severe acute pancreatitis (NSAP, n = 135) and severe acute pancreatitis (SAP, n = 65) based on final clinical diagnosis. Differences in CTSI, general clinical features, and laboratory indicators between the two groups were compared. The LASSO regression model was used to select variables that might affect the severity of AP, and these variables were analyzed using multivariate logistic regression. A nomogram model was constructed using R software, and its AUC value was calculated. The accuracy and practicality of the model were evaluated using calibration curves, Hosmer-Lemeshow test, and decision curve analysis (DCA), with internal validation performed using the bootstrap method. Finally, 60 AP patients treated in the same hospital from July 2022 to December 2023 were selected for external validation.LASSO regression identified CTSI, BUN, D-D, NLR, and Ascites as five predictive factors. Unconditional binary logistic regression analysis showed that CTSI (OR = 2.141, 95%CI:1.369-3.504), BUN (OR = 1.378, 95%CI:1.026-1.959), NLR (OR = 1.370, 95%CI:1.016-1.906), D-D (OR = 1.500, 95%CI:1.112-2.110), and Ascites (OR = 5.517, 95%CI:1.217-2.993) were independent factors influencing SAP. The established prediction model had a C-index of 0.962, indicating high accuracy. Calibration curves demonstrated good consistency between predicted survival rates and actual survival rates. The C-indexes for internal and external validation were 0.935 and 0.901, respectively, with calibration curves close to the ideal line.The model based on CTSI and clinical indicators can effectively predict the severity of AP, providing a scientific basis for clinical decision-making by physicians.

A heterogeneous graph-based semi-supervised learning framework for access control decision-making

For modern information systems, robust access control mechanisms are vital in safeguarding data integrity and ensuring the entire system’s security. This paper proposes a novel semi-supervised learning framework that leverages heterogeneous graph neural network-based embedding to encapsulate both the intricate relationships within the organizational structure and interactions between users and resources. Unlike existing methods focusing solely on individual user and resource attributes, our approach embeds organizational and operational interrelationships into the hidden layer node embeddings. These embeddings are learned from a self-supervised link prediction task based on a constructed access control heterogeneous graph via a heterogeneous graph neural network. Subsequently, the learned node embeddings, along with the original node features, serve as inputs for a supervised access control decision-making task, facilitating the construction of a machine-learning access control model. Experimental results on the open-sourced Amazon access control dataset demonstrate that our proposed framework outperforms models using original or manually extracted graph-based features from previous works. The prepossessed data and codes are available on GitHub,facilitating reproducibility and further research endeavors.

Construction and application of economic resilience evaluation model for megacities.

Economic resilience provides a new perspective for megacities to achieve sustainable development when facing multiple shocks, and its accurate evaluation is an essential prerequisite for optimizing urban governance. There are currently no generally accepted methods for empirical evaluation or measuring economic resilience, and the present study aims to contribute to in both the research field and methodology. The present study sets dimensions and indicators based on economic resilience's theoretical and empirical research and used Decision Making Trial and Evaluation Laboratory (DEMATEL) and Interactive Structural Modeling (ISM) methods to exclude the effect indicators and divide the indicator hierarchy, respectively. Subsequently, the present study conducts model validation using Chinese megacities as a case study. The game theory weighting method, which combines the Analytic Hierarchy Process (AHP) and Entropy methods, is used to calculate indicator weights, and the VIKOR (VIseKriterijumska Optimizacija i KOmpromisno Resenje) method is used to evaluate and compare economic resilience of megacities. The research findings indicate that the evaluation model constructed in the present study included 15 indicators (after excluding three effect indicators) divided into four levels. After merging the levels, they correspond to three dimensions: resistance, recoverability, and adaptability. In addition, using Chinese megacities as a case study, the evaluation results found that Beijing, Shanghai, and Shenzhen have high economic resilience, Tianjin and Guangzhou have moderate economic resilience, Chengdu has low economic resilience, and Chongqing has the lowest economic resilience. This result is consistent with previous studies and verifies the model's effectiveness. The present study also found that megacities with lower levels of economic resilience exhibit a more significant upward trend, as well as the highest and higher proportion of economic resilience in Chinese megacities depending on time passes, indicating that megacities' economic resilience is weakening. The evaluation result obtained in the present study is more specific, precise, and focused on depicting the distribution differences and development trends of economic resilience at the urban level.

Analysis of employee diligence and mining of behavioral patterns based on portrait portrayal

With the deepening of enterprise digital construction, the portrait portrayal based on employee behaviors has gradually become a research focus. Currently, the employee's portrait portrayal mostly has the problems of simple means, low efficiency, limited solving ability, etc., making the results more one-sided. Therefore, a data mining-based employee portrait portrayal model is proposed. The content of employee portrait portrayal is deeply analyzed, and the overall framework of the model is designed. A diligence analysis model (DAM) based on improved GAN is constructed, and the diligence evaluation of employees is clarified to realize the diligence evaluation. The results of diligence analysis of DAM have high accuracy (80.39%) and outperform SA (70.24%), K-means (51.79%) and GAN (67.25%). The Kappa coefficient of DAM reaches 0.7384, which is highly consistent and higher than SA (0.6075), K-means (0.3711) and GAN (0.5661). The Local Outlier Factor (LOF) and Isolation Forest (IF) are used to detect abnormal behaviors on the employees, and mine the abnormal behavior patterns on different granularity time. The LSTM model (Att-LSTM) based on the attention mechanism is used to complete the prediction of employees' software usage behaviors, and analyze and summarize the characteristics of employee's behaviors from multiple perspectives. Att-LSTM predicts the best with an RMSE of 0.82983, which is better than LSTM (0.90833) and SA (0.97767); AM-LSTM has a MAPE of 0.80323, which is better than LSTM (0.86233) and SA (0.92223). The results show that the data mining-based employee portrait portrayal method can better solve the problem of enterprise employees' digital construction, and provide a new way of thinking for the construction of enterprise-level employees' digital portrait model and the analysis of employee behavior.