One of the standard criteria for expressing the relationship between two random variables is the correlation coefficient. Correlation between variables shows that changing the value of one variable leads to changing another variable in a certain direction. It is also possible to use the value of one variable to predict the value of another. In statistics, the correlation coefficient measures the direction and strength of the tendency to change. In machine learning, the correlation coefficient is known as a measure of classification quality. In fact, as a starting step for classification, the correlation between different samples should be estimated using a specific method. There are various methods to estimate the correlation of different data types, which have disadvantages such as low accuracy or high computational time. One of the methods that can overcome these problems, due to its high capability in modeling correlation between samples is graphical modeling. In this research, a new covariance model based on graph theory and graph neural network for estimating the correlation between samples is presented. The results show the improvement of the proposed model in accuracy, sensitivity, precision, F-Micro, F-Macro and statistical tests compared to Pearson and cosine methods.

Robots have become integral to modern society, taking over both complex and routine human tasks. Recent advancements in depth camera technology have propelled computer vision-based robotics into a prominent field of research. Many robotic tasks—such as picking up, carrying, and utilizing tools or objects—begin with an initial grasping step. Vision-based grasping requires the precise identification of grasp locations on objects, making the segmentation of objects into meaningful components a crucial stage in robotic grasping. In this paper, we present a system designed to detect the graspable parts of objects for a specific task. Recognizing that everyday household items are typically grasped at certain sections for carrying, we created a database of these objects and their corresponding graspable parts. Building on the success of the Dynamic Graph CNN (DGCNN) network in segmenting object components, we enhanced this network to detect the graspable areas of objects. The enhanced network was trained on the compiled database, and the visual results, along with the obtained Intersection over Union (IoU) metrics, demonstrate its success in detecting graspable regions. It achieved a grand mean IoU (gmIoU) of 92.57% across all classes, outperforming established networks such as PointNet++ in part segmentation for this dataset. Furthermore, statistical analysis using analysis of variance (ANOVA) and T-test validates the superiority of our method.

Introduction and objectives: In today's world, accessibility and security of banking services for all members of society, particularly vulnerable groups such as the blind, are of utmost importance. With the increasing significance of digital banking, identifying and assessing risks related to accessibility and security of banking services for the blind has become a fundamental priority. This research aims to identify, evaluate, and prioritize the main risks associated with providing banking services to the blind and propose solutions to mitigate these risks. The goal is to improve the infrastructure and technologies used to significantly facilitate access to banking services for the blind. This study employs a combination of two methods: fuzzy Failure Mode and Effects Analysis (FMEA) and Graph Neural Networks (GNN), to more accurately and comprehensively identify the relationships and interactions among risks. Methods: This study was conducted in two main stages. In the first stage, the fuzzy FMEA method was used to identify and evaluate risks. Due to its capability to work with fuzzy numbers, this method is particularly suitable for analyzing the criteria of severity, occurrence, and detectability of risks under conditions of uncertainty. After collecting experts' opinions, these criteria were defuzzified into crisp values, and the risks were prioritized. The second stage involved applying the Graph Neural Network (GNN) method to model and analyze the complex dependencies and interrelationships among the risks. GNN, as a powerful machine learning tool, enables the examination of interdependencies among different criteria and nodes. The research data were gathered through surveys conducted with 12 experts in banking and specialized services for the blind. Each expert was presented with a questionnaire containing various pairs of risk criteria and was asked to assign a score between 0 and 4 to each pair. To reduce the impact of individual opinions and achieve a comprehensive assessment, the average scores given by the experts were used as the final weights of the relationships among the criteria in the graph. Findings: The results of the fuzzy FMEA analysis revealed that "physical access," "economic inequalities," "digital divide," and "technological barriers" are among the most significant risks to the accessibility of banking services for the blind. The non-fuzzy Risk Priority Number (RPN) results indicated that the risks "physical access" and "economic inequalities" require the highest priority attention and demand special focus. The GNN analysis confirmed that some risks, such as physical access and technological barriers, have complex and mutual effects on other risks and play a crucial role in the network of relationships among criteria. Specifically, the criteria "economic inequalities" and "technological barriers" were identified as key influencing factors within the graph network. Addressing these risks can significantly improve the accessibility and banking experience for the blind. Furthermore, the findings emphasized that focusing solely on economic and technological aspects is insufficient; the interactions among these criteria must also be considered. Conclusion: Enhancing access to banking services for the blind requires a multifaceted approach that simultaneously focuses on improving physical infrastructure, reducing economic inequalities, raising awareness and providing training on banking technologies, and strengthening information security. The findings of this study demonstrate that integrating fuzzy FMEA and GNN can effectively identify interactions and prioritize risks more accurately, providing a foundation for designing more comprehensive and impactful solutions to improve the accessibility of banking services for the blind. It is recommended that banks and financial institutions utilize these findings to implement inclusive solutions that enhance accessibility and user experience for the blind. Such efforts can ultimately increase customer satisfaction and trust, improving the credibility and social responsibility of banks.

Tax evasion based on related party transactions is a new strategy in tax evasion that is carried out through legal transactions, such as transactions between a group of companies that have heterogeneous, complex, and hidden interaction relationships for tax evasion. Existing studies cannot effectively identify tax evasion behaviors of related parties because the machine learning-based audit method can detect the abnormal financial status of individuals with high accuracy and efficiency. However, it is helpless when faced with heterogeneous, complex, and hidden interaction relationships and cannot identify tax evasion groups with related party transactions. The hybrid of graph mining and deep neural network approaches has the ability to detect anomalies in complex organizational structures. In this study, 1,780 companies with related party transactions, including 523 companies located in free trade zones and 1,257 companies located outside free trade zones, which have a common board member and economic activity of production or trade, were selected. In this study, financial and tax data from tax returns and the systems of the Iranian Tax Administration from 2016 to 2019 were used. This study is practical in terms of purpose. Python software and the NetworkX package were used to estimate the model. To predict tax evasion in related party transactions, three algorithms were used: Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Multilayer Perceptron Neural Network (MLP) in deep mode. To identify suspicious groups, three steps were taken; first: detecting tax rate differences, matching the topological pattern, and identifying tax burden anomalies; second: experimental tests based on data from 16,756 related party transaction purchases and sales in the country; third: estimating the coefficients and the relationship between the topological pattern in the two cases of profit retention and profit transfer based on the graph mining approach and deep neural network. The results show that both profit retention and profit shifting exist in tax evasion of related party transactions. However, based on the results, the intensity of the profit retention relationship in tax evasion of related party transactions is stronger than the profit shifting relationship. Based on the results, the graph mining approach was more accurate than the logit, probit, and linear probability models.

The increase of cameras nowadays, and the power of the media in people's lives lead to a staggering amount of video data. It is certain that a method to process this large volume of videos quickly and optimally becomes especially important. With the help of video summarization, this task is achieved and the film is summarized into a series of short but meaningful frames or clips. This study tried to cluster the data by an algorithm (K-Medoids) and then with the help of a convolutional graph attention network, temporal and graph separation is done, then in the next step with the connection rejection method, noises and duplicates are removed, and finally summarization is done by merging the results obtained from two different graphical and temporal steps. The results were analyzed qualitatively and quantitatively on three datasets SumMe, TVSum, and OpenCv. In the qualitative method, an average of 88% accuracy rate in summarization and 31% error rate was achieved, which is one of the highest accuracy rates compared to other methods. In quantitative evaluation, the proposed method has a higher efficiency than the existing methods.