Federated learning has been used extensively in business inno-vation scenarios in various industries.This research adopts the federated learning approach for the first time to address the issue of bank-enterprise info...Federated learning has been used extensively in business inno-vation scenarios in various industries.This research adopts the federated learning approach for the first time to address the issue of bank-enterprise information asymmetry in the credit assessment scenario.First,this research designs a credit risk assessment model based on federated learning and feature selection for micro and small enterprises(MSEs)using multi-dimensional enterprise data and multi-perspective enterprise information.The proposed model includes four main processes:namely encrypted entity alignment,hybrid feature selection,secure multi-party computation,and global model updating.Secondly,a two-step feature selection algorithm based on wrapper and filter is designed to construct the optimal feature set in multi-source heterogeneous data,which can provide excellent accuracy and interpretability.In addition,a local update screening strategy is proposed to select trustworthy model parameters for aggregation each time to ensure the quality of the global model.The results of the study show that the model error rate is reduced by 6.22%and the recall rate is improved by 11.03%compared to the algorithms commonly used in credit risk research,significantly improving the ability to identify defaulters.Finally,the business operations of commercial banks are used to confirm the potential of the proposed model for real-world implementation.展开更多
In this paper, we study edge detection or segmentation, which is recognized as a rudiment innovation as it can evaluate sharpness and analyze object boundaries. That’s the reason it has been an influential figure in ...In this paper, we study edge detection or segmentation, which is recognized as a rudiment innovation as it can evaluate sharpness and analyze object boundaries. That’s the reason it has been an influential figure in the image-processing era. Because of this, it has a significant influence in the age of image processing. On the other hand, edge detection is the process of dividing an image into discontinuous regions. It specifies the intensity shift connected to the image’s edge. There are several methods for detecting edges. Four edge identification methods on satellite images and satellite images affected by Gaussian noise were examined. Known edge detection technologies such as Canny, Prewitt, Scharr, and Robert operators are included in this study. Additionally, the key feature of an image for evaluating its quality is the Image Quality Assessment (IQA) measure. We primarily take into account SSIM, MSE, PSNR, and RMSE when assessing image quality. Experimental validation has been obtained for the application of the Canny and Prewitt algorithms to the satellite dataset. However, when the Gaussian Noise effect is added to the same dataset, clever edge detection performs better.展开更多
Compared to high-resolution digital-toanalog converters(DACs), deploying 1-bit DACs requires much less hardware complexity for a massive multi-user multiple-input multiple-output(MUMIMO) system. However, the feasible ...Compared to high-resolution digital-toanalog converters(DACs), deploying 1-bit DACs requires much less hardware complexity for a massive multi-user multiple-input multiple-output(MUMIMO) system. However, the feasible domain of a1-bit transmitting signal is non-continuous, and thus it is more challenging to exploit multi-user interference(MUI) by precoding. In this paper, to improve symbol decision accuracy, we investigate MUI exploitation 1-bit precoding methods for massive MU-MIMO systems under QAM modulations. Because MUIs may be constructive or destructive, we define a modified mean square error(MSE) metric for QAM constellations to jointly evaluate the effect of both MUIs and noise. Then, we model the 1-bit precoding optimization problems to minimize the sum modified MSE or the maximum modified MSE, where both the transmitting vector and receiving processing factor are optimization variables. Based on whether the receiving processing factor remains constant during the whole transmission block, two scenarios are taken into consideration. Referring to existing interference exploitation 1-bit precoding methods, we design efficient algorithms to solve the two modified MSE based problems.Compared to existing 1-bit precoding methods, our proposed methods provide better bit error rate performance, especially in more practical scenario Ⅱ(constant receiving processing factor in one block).展开更多
Due to the rapid development of logistic industry, transportation cost is also increasing, and finding trends in transportation activities will impact positively in investment in transportation infrastructure. There i...Due to the rapid development of logistic industry, transportation cost is also increasing, and finding trends in transportation activities will impact positively in investment in transportation infrastructure. There is limited literature and data-driven analysis about trends in transportation mode. This thesis delves into the operational challenges of vehicle performance management within logistics clusters, a critical aspect of efficient supply chain operations. It aims to address the issues faced by logistics organizations in optimizing their vehicle fleets’ performance, essential for seamless logistics operations. The study’s core design involves the development of a predictive logistics model based on regression, focused on forecasting, and evaluating vehicle performance in logistics clusters. It encompasses a comprehensive literature review, research methodology, data sources, variables, feature engineering, and model training and evaluation and F-test analysis was done to identify and verify the relationships between attributes and the target variable. The findings highlight the model’s efficacy, with a low mean squared error (MSE) value of 3.42, indicating its accuracy in predicting performance metrics. The high R-squared (R2) score of 0.921 emphasizes its ability to capture relationships between input characteristics and performance metrics. The model’s training and testing accuracy further attest to its reliability and generalization capabilities. In interpretation, this research underscores the practical significance of the findings. The regression-based model provides a practical solution for the logistics industry, enabling informed decisions regarding resource allocation, maintenance planning, and delivery route optimization. This contributes to enhanced overall logistics performance and customer service. By addressing performance gaps and embracing modern logistics technologies, the study supports the ongoing evolution of vehicle performance management in logistics clusters, fostering increased competitiveness and sustainability in the logistics sector.展开更多
.Abstracting eye models from MRI images is critical in advancing medical imaging, particularly for clinical diagnostics. Current methods often struggle with accuracy and efficiency, highlighting a gap this research ai....Abstracting eye models from MRI images is critical in advancing medical imaging, particularly for clinical diagnostics. Current methods often struggle with accuracy and efficiency, highlighting a gap this research aims to fill. This study investigates the application of machine learning methods, focusing on the U-net-based deep learning framework, to improve the accuracy of eye model extraction. The objectives include fitting measured eye data to models such as the Ellipsoid model, evaluating automated segmentation tools, and assessing the usability of machine learning-based extractions in clinical scenarios. We employed point cloud data of 202,872 points to fit eye models using ellipsoid, non-linear, and spherical fitting techniques. The fitting processes were optimized to ensure precision and reliability. We compared the performance of these models using mean squared error (MSE) as the primary metric. The non-linear model emerged as the most accurate, with a significantly lower MSE (1.186562) compared to the ellipsoid (781.0542) and spherical models. This finding indicates that the non-linear model provides a more detailed and precise representation of the eye’s geometry. These results suggest that machine learning methods, particularly non-linear models, can significantly enhance the accuracy and usability of eye model extraction in clinical diagnostics, offering a robust framework for future advancements in medical imaging.展开更多
基金funded by the State Grid Jiangsu Electric Power Company(Grant No.JS2020112)the National Natural Science Foundation of China(Grant No.62272236).
文摘Federated learning has been used extensively in business inno-vation scenarios in various industries.This research adopts the federated learning approach for the first time to address the issue of bank-enterprise information asymmetry in the credit assessment scenario.First,this research designs a credit risk assessment model based on federated learning and feature selection for micro and small enterprises(MSEs)using multi-dimensional enterprise data and multi-perspective enterprise information.The proposed model includes four main processes:namely encrypted entity alignment,hybrid feature selection,secure multi-party computation,and global model updating.Secondly,a two-step feature selection algorithm based on wrapper and filter is designed to construct the optimal feature set in multi-source heterogeneous data,which can provide excellent accuracy and interpretability.In addition,a local update screening strategy is proposed to select trustworthy model parameters for aggregation each time to ensure the quality of the global model.The results of the study show that the model error rate is reduced by 6.22%and the recall rate is improved by 11.03%compared to the algorithms commonly used in credit risk research,significantly improving the ability to identify defaulters.Finally,the business operations of commercial banks are used to confirm the potential of the proposed model for real-world implementation.
文摘In this paper, we study edge detection or segmentation, which is recognized as a rudiment innovation as it can evaluate sharpness and analyze object boundaries. That’s the reason it has been an influential figure in the image-processing era. Because of this, it has a significant influence in the age of image processing. On the other hand, edge detection is the process of dividing an image into discontinuous regions. It specifies the intensity shift connected to the image’s edge. There are several methods for detecting edges. Four edge identification methods on satellite images and satellite images affected by Gaussian noise were examined. Known edge detection technologies such as Canny, Prewitt, Scharr, and Robert operators are included in this study. Additionally, the key feature of an image for evaluating its quality is the Image Quality Assessment (IQA) measure. We primarily take into account SSIM, MSE, PSNR, and RMSE when assessing image quality. Experimental validation has been obtained for the application of the Canny and Prewitt algorithms to the satellite dataset. However, when the Gaussian Noise effect is added to the same dataset, clever edge detection performs better.
文摘Compared to high-resolution digital-toanalog converters(DACs), deploying 1-bit DACs requires much less hardware complexity for a massive multi-user multiple-input multiple-output(MUMIMO) system. However, the feasible domain of a1-bit transmitting signal is non-continuous, and thus it is more challenging to exploit multi-user interference(MUI) by precoding. In this paper, to improve symbol decision accuracy, we investigate MUI exploitation 1-bit precoding methods for massive MU-MIMO systems under QAM modulations. Because MUIs may be constructive or destructive, we define a modified mean square error(MSE) metric for QAM constellations to jointly evaluate the effect of both MUIs and noise. Then, we model the 1-bit precoding optimization problems to minimize the sum modified MSE or the maximum modified MSE, where both the transmitting vector and receiving processing factor are optimization variables. Based on whether the receiving processing factor remains constant during the whole transmission block, two scenarios are taken into consideration. Referring to existing interference exploitation 1-bit precoding methods, we design efficient algorithms to solve the two modified MSE based problems.Compared to existing 1-bit precoding methods, our proposed methods provide better bit error rate performance, especially in more practical scenario Ⅱ(constant receiving processing factor in one block).
文摘Due to the rapid development of logistic industry, transportation cost is also increasing, and finding trends in transportation activities will impact positively in investment in transportation infrastructure. There is limited literature and data-driven analysis about trends in transportation mode. This thesis delves into the operational challenges of vehicle performance management within logistics clusters, a critical aspect of efficient supply chain operations. It aims to address the issues faced by logistics organizations in optimizing their vehicle fleets’ performance, essential for seamless logistics operations. The study’s core design involves the development of a predictive logistics model based on regression, focused on forecasting, and evaluating vehicle performance in logistics clusters. It encompasses a comprehensive literature review, research methodology, data sources, variables, feature engineering, and model training and evaluation and F-test analysis was done to identify and verify the relationships between attributes and the target variable. The findings highlight the model’s efficacy, with a low mean squared error (MSE) value of 3.42, indicating its accuracy in predicting performance metrics. The high R-squared (R2) score of 0.921 emphasizes its ability to capture relationships between input characteristics and performance metrics. The model’s training and testing accuracy further attest to its reliability and generalization capabilities. In interpretation, this research underscores the practical significance of the findings. The regression-based model provides a practical solution for the logistics industry, enabling informed decisions regarding resource allocation, maintenance planning, and delivery route optimization. This contributes to enhanced overall logistics performance and customer service. By addressing performance gaps and embracing modern logistics technologies, the study supports the ongoing evolution of vehicle performance management in logistics clusters, fostering increased competitiveness and sustainability in the logistics sector.
文摘.Abstracting eye models from MRI images is critical in advancing medical imaging, particularly for clinical diagnostics. Current methods often struggle with accuracy and efficiency, highlighting a gap this research aims to fill. This study investigates the application of machine learning methods, focusing on the U-net-based deep learning framework, to improve the accuracy of eye model extraction. The objectives include fitting measured eye data to models such as the Ellipsoid model, evaluating automated segmentation tools, and assessing the usability of machine learning-based extractions in clinical scenarios. We employed point cloud data of 202,872 points to fit eye models using ellipsoid, non-linear, and spherical fitting techniques. The fitting processes were optimized to ensure precision and reliability. We compared the performance of these models using mean squared error (MSE) as the primary metric. The non-linear model emerged as the most accurate, with a significantly lower MSE (1.186562) compared to the ellipsoid (781.0542) and spherical models. This finding indicates that the non-linear model provides a more detailed and precise representation of the eye’s geometry. These results suggest that machine learning methods, particularly non-linear models, can significantly enhance the accuracy and usability of eye model extraction in clinical diagnostics, offering a robust framework for future advancements in medical imaging.
