نشریه محاسبات نرم
سال دهم شماره 2 (پیاپی 20، پاییز و زمستان 1400)

In recent years, the issue of credit rating and identification of good and bad customers have received a lot of attention from banks. Granting facilities to well-accounted customers and avoiding granting facilities to badly accounted customers, which leads to reducing bank arrears, are always the major concerns of bank managers, which are not out of reach with the help of an efficient and good credit rating system. This paper presents a new ensemble model based on support vector machine algorithm for the credit rating of bank customers. First, the data set is divided into several subsets by a bootstrap approach. The support vector machine algorithm is then applied to each subset and several models are formed. At the end, voting is done between the models, and the final model is obtained. In order to show the accuracy of the ensemble model, the credit data of Pasargad Bank’s costumers including 2218 instances of credit applicants, each instance contains 14 explanatory attribute, are evaluated using the proposed method. Based on different criteria, the results obtained on the data of the Pasargad Bank’s costumers confirm the superiority of the ensemble support vector machine method over the usual support vector machine method and the random forest method. The type II error (indicates the proportion of bad applicants who are wrongly predicted to be good applicants) of the proposed ensemble method with linear core is 17% less than the usual support vector machine method and 15% less than the random forest method.

The power industry is facing problems such as fossil fuel reduction, greenhouse gas emissions and the lack of automated analysis of the power grid. To meet the existing challenges, creating a new infrastructure is essential. The concept of smart grids aims to update existing power grids by integrating advanced communication infrastructure with the existing grid, making power grids more reliable, efficient, and environmentally friendly. Integration of power networks and communication networks, while it has many benefits, it has disadvantages in system security and protection. In this paper, after familiarizing with the smart grid communication infrastructure, their features and applications, the existing threats against the security of communication networks are introduced, and suitable actions are discussed, including prevention, detection, and counteract against security risks. At the security risks prevention stage, solutions such as encryption and blockchain have been studied and reviewed with the purpose of maintaining access to network information for authorized users. In the field of identifying attacks in the smart grid due to the need for a smart and fast solution, deep learning as one of the smart, dynamic and powerful tools with its applications in improving smart grid security has been investigated.

One of the most common CPU scheduling algorithms is Round Robin (RR). In RR, a time quantum represents the maximum time that a process can have the processor, and the processor is allocated to the processes of ready queue in circular order. The size of time quantum has a significant effect on the performance of RR, so that if it is small, due to the increase in the number of context switches and the resulting overhead, the CPU utilization will decrease. In contrast, if the time quantum is large, the average response time of the processes increases, which makes the use of RR in interactive applications inefficient. The purpose of this paper is to provide an efficient method for determining dynamic time quantum using machine learning. Initially, a training set including the number of processes and the maximum, minimum, average and median of their burst time and optimal time quantum as the class has been developed. Then, by training machine learning classifiers on this set, the optimal time quantum for new examples is predicted. The experimental results show that in general, the proposed method has a better performance compared to other methods based on the performance measures of scheduling algorithms. For instance, compared to the genetic algorithm, which has the best performance among the existing methods, the average waiting time and the average number of context switches of the proposed method improved by 12 milliseconds and 1.76 units, respectively, and the average turnaround time is increased about 2 milliseconds.

Augmented reality is one of the new technologies in recent decade that has appeared especially in mobile devices. In augmented reality, virtual objects are placed in a real environment, next to other real objects. Another reason for the importance of augmented reality is the implementation of augmented reality applications on mobile and personal devices such as smartphones and tablets. The integration of augmented reality with a numerical simulation method such as computational fluid dynamics provides a cognitive, scientific and efficient tool for expert and non-professional users to analyze practical problems. In addition, using computational fluid dynamics techniques and an augmented reality-based system, engineering analysis and simulation results are obtained directly on real-world objects, resulting in a better understanding and relationship between the simulation results and the world. It becomes real. In this research, the combination of augmented reality and computational fluid dynamics method has been used to implement an Android software to simulate the thermal changes of solids with different materials over time.

As coronavirus disease progressed, most societies experienced an increase in divorce rates due to dissatisfaction, incompatibility, and turmoil in couples' relationships. One of the effective factors in a successful marriage is the realistic expectation of marriage that examining the factors of a successful marriage is an important step to achieve this aim. In this regard, we investigated the effective factors in a successful marriage or divorce of couples using machine learning algorithms. To predict the type of couple relationship, this study designed a standard questionnaire with 54 questions based on Guttman's couple therapy questionnaire in the Kaggle database. The purpose of this questionnaire is to determine the survival of the relationship using the analysis of 170 couples' responses. The designed questionnaire was distributed among 33 Iranian families and the type of relationship between couples was predicted using machine learning algorithms. Also, the effective features of the relationship between couples in Iranian culture in this study were extracted with the help of Pearson correlation coefficient and with the effective features in another study that was performed on people with different cultures; Compared and the results showed that culture change can change some of the effective characteristics. Experimental results demonstrated that the effective factors in predicting a person's divorce or marriage are: 1- Intimacy between couples and 2- Silence and suppression of anger in discussions and disputes. Among the examined machine learning algorithms, the decision tree and naive Bayes have shown the highest efficiency with 100% accuracy.

In the recent years, the issue of increasing software security has become to extensive research in the software produced process. One of main features secure systems engineering is identify security requirement to protect assets. The cost, time and effecting features of each method of identify security requirements has become important for providing new methods for implementing security in software systems. These security methods implement in order to increase confidence from systems and protection of organization of systems in software production cycles. Engineers must be sure that software produced is secure against potential threats and attackers.Identity, classification, prioritize and measuring security requirements in terms of privacy, software structure and other issues are part of the concerns of software developers. In this study, a selection of new methods of security requirements and their security factors from 2017 to 2021 are collected and compared the factors of each method in terms of their maximum repetition. This study introduces how new methods work to developer, has become better chosen and has been selected most appropriate method. It also helps people who are looking for recent research in the field of security requirements.

Today, the automatic license plate recognition system is of special importance due to its various applications. In general, this system consists of four basic stages: capture an image of the car, finding the place of the license plate from the image, separating the characters from the license plate and recognizing the characters. In this paper, a set of morphological operations as well as vertical image histogram processing is presented to improve the accuracy of the plate license recognition. Indeed, morphological operations is used to determine candidate plate license and vertical histogram of the image is incorporate to remove false candidate locations. The simulation results on two datasets color and infrared images show that the proposed method is competitive with other methods.

Image segmentation is the process of dividing a digital image into several parts. The segmentation goal is to simplify, or change the representation of an image into something that is both more meaningful and easier to analyze. Thresholding methods with much less complexity are still widely used compared to modern methods based on deep learning. In this paper, a new optimal multi-level thresholding algorithm for histogram-based segmentation of images is presented. The proposed algorithm compared to Particle Swarm Optimization Algorithm (PSO) and an improved version of the PSO based on multi-agent fuzzy which is called MAFPSO. In the proposed algorithm, the selection of image thresholding is done using the recently introduced buzzard optimization algorithm (BUZO). In the BUZO algorithm, the process of exploration and exploitation is achieved by defining several types of buzzard with different abilities. Multilevel segmentation is performed-using entropy as a fitness function for BUZO. Comparing the performance of BUZO algorithm with MAFPSO, and PSO for several benchmark images show 8 percent average improvement for fitness function. The quality of segmented images shows 3% in average improvement for 2-level segmented image, and shows 12% in average improvement for 5-level segmented images.

In this paper, we introduce Legendre multiwavelet functions and use them as a set of base functions to approximate the solution of a parabolic differential equation with an unknown time-dependent coefficient in an inverse problem. Using the expansion formula of a known function in terms of the Legendre multiwavelet base, we define integral and product operational matrices from a general point of view. With the help of these matrices, we transform the problem into a system of algebraic equations. By solving the obtained system of algebraic equations using the existing optimization algorithms, we provide an approximation for the solution of the problem in the form of its expansion in terms of the Legendre multiwavelet base. In addition to expressing the algorithm of the proposed numerical method, we perform the proposed method on two examples and report its numerical results. We also compare the results of the proposed method with the results reported from other methods.