INTERNATIONAL CONFERENCE ON WEB RESEARCH | Year:2022 | دوره:8 |Papers Count:53

In recent years, with the increasing advancement of artificial intelligence and machine learning, huge changes have occurred in the field of vehicles and self-driving cars. Recognition and interpretation of urban traffic signs, by machine vision systems, improve the safety of driving operations as one of the basic principles of self-driving vehicles. Due to the high interaction of self-driving vehicles with traffic signs during the movement, creating a system with high accuracy for interpretation and an immediate decision is a big challenge. In this research, with the use of convolutional neural networks, a system is designed that can recognize Iranian traffic signs. By applying the transfer learning approach, we train our model with a new collection of traffic signs images that reaches a high accuracy in optimal conditions.

In a short period of time, Bitcoin digital currency has been able to transform businesses and the global economy to a great extent. The growth and survival of any business in today's competitive world depends on economic, political, social, technological, legal and environmental factors. Successful development of cryptocurrencies in businesses requires careful analysis of these factors and the achievement of a strategic application plan. In this research, the internal factors (strengths and weaknesses) and external factors (opportunities and threats) affecting the use of Bitcoin in businesses are identified with the help of existing reports, articles and documents and with the help of qualitative and quantitative strategic analysis methods SWOT and QSPM its development strategies are extracted. It is hoped that the proposed strategies of this study achieve the success of businesses in using bitcoin digital currency.

The widespread development of wireless networks and cloud computing has provided numerous benefits to individuals in the community. These networks have applications in the fields of health and medicine, including medical wireless sensor networks. One of the concerns in these networks is security. In the context of activities in this area, the issue of security and privacy remains one of the challenges, and only a small number of schemes have been able to achieve good security from the proposed schemes. Sureshkumar et al. recently presented an authentication scheme for IoT-based smart medical services, claiming that it is resistant to known attacks. This paper demonstrates that Sureshkumar et al. authentication's scheme is vulnerable to traceability, integrity contradiction, and de-synchronization attacks. This paper also examines the security of Sureshkumar et al. authentication's protocol using the Scyther tool, which verifies the accuracy of the presented attacks. In this paper, we also make recommendations to improve Sureshkumar et al., so that this protocol can provide complete security against all active and passive attacks, particularly the attacks presented in this paper.

Meta‑ Heuristic algorithms are optimization techniques that provide the optimal solution through processes of repeated exploration and exploitation of the entire search space. Feature selection is also an important and prominent process in the field of machine learning that reduces data dimensions. This paper examines and compares nature-inspired meta-heuristic algorithms for feature selection to increase the accuracy of software fault prediction. Researchers can not easily select meta-heuristic algorithms as a suitable method for their research due to their great variety and multiplicity. In this paper, by describing the feature selection techniques and its methods, the application of meta-heuristic algorithms in different fields, such as swarm intelligence and binary methods of these algorithms has been investigated. Also, by introducing 18 meta-heuristic algorithms in 6 different categories and evaluating each of them, a suitable analysis has been provided to researchers so that they can easily and with the highest efficiency choose the appropriate algorithm and method of their work. In the papers presented so far, meta-heuristic algorithms have been studied from only one aspect, while in this article, while studying different types of research, they have tried to study and evaluate them from different aspects.

Education mainly contributes to the spread of knowledge and plays an effective role in eliminating ignorance and illiteracy in the world. Through education, some countries have been able to achieve wealth and civilization, which made them one of the developed countries. Education is a need and a necessity for all human beings, to achieve the dreams and ambition of studying and acquiring knowledge in various educational institutions that are not available in a particular country, based on this principle, so-called distance learning appeared, but today, for the continuation of the educational process under the critical conditions and various crises especially corona crisis, distance learning has formed the rescue gate for the continuation of the education process in the whole world. This study aims to analysis and illustrated that distance learning from a global and local perspective as ideal solve for continue the educational process through the various crisis and the donor for giving and providing chances for all the learners.

In Wireless Body Area Networks (WBANs), the sensor energy is limited. Due to dynamic and huge data exchange, sending data consumes the most sensor energy. The best solution to solve this problem is to use data compression methods. The Compressed Sensing (CS) method is among the popular methods for compressing data in WBANs. The problem with this method is does not work well when the data set is not sparse. In this paper, to solve this problem, two versions of Block Sparse Bayesian Learning (BSBL) Bound-Optimization (BSBL-BO), and Expectation-Maximization (BSBL-EM) are used to compress and recover the Arterial Blood Pressure systolic (ABPsys) and Respiration signals. These signals are adapted and reshaped to the BSBL environment as an input dataset and then compressed. The phi matrix is created compatible with ABPsys and Respiration signals and obtained 98% similarity with the original signal after restoration. According to the results, the similarity of ABPsys and Respiration signals after recovery by BSBL-BO is higher than the BSBL-EM method. BSBL-BO is faster at signal recovery than BSBL-EM. The amount of residual energy is compared between the two CS methods, DCT, as dictionary matrix in CS using the BSBL versions, and the DCT without the BSBL and DCT with BSBL performs better than alone DCT.

Public acceptance of social networks has made the analysis of these networks essential. Event detection in these networks including Twitter is one of the most momentous subjects in the field of natural language processing and text mining. In this paper, we investigated how to link popular social media topics and news stories using transformer models and neural networks. Accordingly, this study consists of two parts: First, detecting popular topics and, second, linking them to the news. Event detection techniques have been applied to detect popular topics, while an event detection method comprises text preprocessing, text embedding using Sentence Transformer, dimension reduction using the UMAP algorithm, and grouping them using the HDBSCAN algorithm. To examine relevance or non-relevance between the news and topics, a single-layer perceptron neural network is applied, in which the output of the model indicates relevance or non-relevance. We have implemented the mentioned parts and have investigated them on a small sampling of two known datasets. The evaluation outcomes reveal that the first part leads to an average improvement of 8% compared to the entity-based methods. Moreover, the results of the second part demonstrate that the used neural network in this study has a better performance comparing several other methods.

Many of real-world social networks, show structural changes over time, so they can be modeled as dynamic graphs. However, most methods in social network analysis, including community detection, are focused on performing on static networks. Therefore, methods of studying community evolution still have room for improvement. In this article, we examine one of the methods introduced in independent community detection and matching approach. It is an approach for tracking dynamic community evolution, but it has the advantage of using methods that have been studied in detail for static networks. Previous studies have examined and compared some of the centralities that can be used in this method. In this study, we examined its performance by using other centralities called betweenness centrality and closeness centrality, and compared them with the usage of social position. Our analysis was performed on a subgraph of the word co-occurrence network, which is a type of bibliometric network, and the results of the algorithm were evaluated by experts. The results shows that betweenness centrality represents more transparent and useful events and using it in community evolution discovery is recommended for small networks.

In recent years, people spend much time on social networks. They use social networks as a place to comment on personal or public events. Thus, a large amount of information is generated and shared daily in these networks. Such a massive amount of information help authorities to accurately and timely monitor and react to events. This unique specification prevents further damages, especially when a crisis occurs. Thus, event detection is attracting considerable interest among social networks research. Since Twitter is one of the most popular social networks that potentially prepare an appropriate bed for event detection, this study has been conducted on Twitter. The main idea of this research is to differentiate among tweets based on some of their features. For this purpose, the proposed methodology applies weights to the three features, including the followers' count, the retweets count, and the user location. The event detection performance is evaluated by scoring potential clusters based on weighting the three mentioned features. The results show that the average execution time and the precision of event detection in the proposed approach have been improved by 27% and 31%, respectively, in comparison to the base method. Another result of this research is detecting more events (including hot events and less important ones) in the presented method.

The proliferation of social networks has made researchers turn to the analysis of these networks. Event detection is one of the important topics in the analysis of social networks, especially Twitter. In this paper, we propose an online graph-based approach, called TedGram, for event detection in Twitter using word embedding techniques and graph partitioning algorithms. In the TedGram model, for each incoming tweet, candidate tweets are gathered from preceding tweets using co-occurrence in entities keywords, and correspondingly the similarity between tweets are computed using the Word Mover’ s Distance (WMD) algorithm and pre-trained word2vec model. In this regard, the TTI (Tweet Tweet Interaction) graph is computed and updated using an online greedy community detection method based on the Barabá si– Albert generative model. Furthermore, we utilize Latent Dirichlet Allocation (LDA) and WMD to combine duplicate communities for detecting and merging duplicate events. Our proposed method is applied to a sample of the Event2012 dataset and is evaluated regarding Precision, Recall, and F-score. The experimental results show that TedGram performs well against the existing methods.

In this paper, a neural network long short term memory hybrid model focusing on stock index forecasting is modeled, presented, and investigated. This model is tuned by artificial intelligence, and neural networks in Python environment. Accordingly, it can perform the prediction with a high accuracy, and near to the real value. Four layers of input layer, hidden layer, attention layer, as well as the output layer set up the proposed hybrid model. The input data, and API-based server connection are performed in input layer. The hidden layer performs the calculations, and measurements. Price value forecasting, and prediction-train graphs done by the attention layer, and output layer, respectively. The proposed system conceptualized the effect of artificial intelligence (AI), and machine learning (ML) on financial markets. Finally, it is concluded this model can be utilized in other wide range of financial applications.

In this paper, the Iran railway network has been analyzed based on structural properties and vulnerabilities. To do this, more than 400 cities have been extracted from Iran railway information and the network has been constructed accordingly. Then, multiple structural properties of the network have been investigated including degree distribution, betweenness, clustering coefficient, and distance distribution. Finally, the network reliability has been studied intensively using random as well as adversarial attacks. According to the calculations, the average degree of the nodes is about 2 and the average shortest path length is about 40. Thus the network would require a structural optimization to improve the economic benefits. Also, according to the attacks made on the nodes, the global efficiency of the network will be dropped more rapidly using the maximum-betweenness attack. Moreover, only small parts of the network could keep their functionality as the size of the giant component is decreased very sharply when less than 20% nodes are removed from the network randomly or intentionally.

Recognizing emotions from text applies to almost every aspect of our daily lives, like improving computer-human interactions, mental health monitoring, recognizing public sentiment about any national, international, or political event. Given the importance of emotion analysis, especially the classification of multi-labeled emotions, this paper is the development of a deep learning-based system that addresses the issue of classifying multi-labeled emotions in texts. To this end, by combining several datasets, we first create a dataset which all samples are multi-labeled, and then, using the long short-term memory recurrent neural network, a new network is designed to detect multiple emotions from the texts. The GloVe and FastText have been used to find semantic, syntactic, and related words. To improve the accuracy of the proposed architecture, we employed the attention property. The comparative results indicated that the proposed model is more accurate than the existing methods.

For most institutions, the World Wide Web has been the main source of information in recent years. Evaluation of a website reveals the overall and actual performance of that website. Identifying a website's strengths and weaknesses is significantly important to any institution. Therefore, in this study, we seek the answer to the question "When evaluating a website, what kind of questionnaires can be used? " The purpose of this article is to give an overview of different questionnaires for the evaluation of web pages. In addition, it explains how these questionnaires can be used. To do this, we used a simple review of the internet to find questionnaires that evaluate the quality of websites.

Today, the growth of the coronavirus as a pandemic and its global expansion is a significant concern in our society and the international community. However, in recent years, many individuals have shifted their major source of news and information to social networks. Consequently, the widespread dissemination of false and misleading information on social media is significant for most politicians. Our effort is not only against COVID-19 but against an "infodemic" as well. To address this, on COVID-19, we have collected and released a labeled dataset of 7, 000 social media postings Persian data, and articles of authentic and false news. Covid 19 fake news has been detected in other languages such as Arabic, English, Chinese, and Hindi. We execute a multi-label task (actual vs. fictitious) on the labeled dataset and compare it to six machine learning baselines: Logistic Regression, Support Vector Machine, Decision Tree, Naive Bayes, K-Nearest Neighbors, and Random Forest. On the test set, the support vector machine gives us the best results, with an 89 percent accuracy rate.

The increasing population has led to a disruption of health status due to the waste management system. Waste management in the form of smart cities has created a kind of concern and is an activity that can provide environmental, health, and social benefits. Waste management is a technique to prevent the accumulation of waste. If it is not managed properly, it will lead to a series of unhealthy and unsanitary conditions in the city. Therefore, to prevent and reduce waste it is needed an intelligent waste management system and one of the newest platforms in this basin is the Internet of Things (IoT) platform. The IoT has revolutionized the waste management system. Different governments are trying to meet this great challenge by creating different IoT-based waste management systems and for this purpose different technologies and sensors such as Bluetooth ultrasonic sensors, RFID tags, etc. are used in their architectures to receive real-time information. Due to the benefit of the use of the IoT in waste management, this article analyzes some of the most recent methods for managing and reducing waste based on the Internet of Things, as well as evaluating various barriers, challenges, and solutions. Also, the characteristics of each method and the advantages and disadvantages are stated.

Identifying influential nodes to influence maximization plays an important role in social networks. Social networks are a type of graph data in which each node represents one person, and each edge represents a relationship between two people. Based on the relationship and interaction among people in social networks, they are influenced by each other, and different users in social networks propagate a large amount of information daily. Thus, it is essential to identify the influential people in spreading information. Identifying influential users is modeled as a problem of influence maximization. This paper proposes a pruning method to identify influential nodes effectively in a smaller graph. Therefore, after extracting the social network data, the edges of the graph are weighted by the edge betweenness centrality measure, and then the edges that weigh less than the average weight are pruned, and the influential nodes in the pruned graph are selected using any centrality measure. To evaluate the proposed pruning method, using the LTM diffusion model, the running time and the average number of activated nodes based on a set of initial active nodes compared to the baseline algorithms based on the centrality algorithm have been reported. The simulation results show a relative improvement in the results obtained.

Covid-19 virus, which is a mutated species of the coronavirus family, was announced as a pandemic by the World Health Organization in March 2020. The death toll from the virus at the beginning of 2022 is close to six million. So far, several methods have been introduced to diagnose the virus, which are mostly difficult to access or require a high cost, and of course, some of these methods are time consuming. Therefore, providing a smart web-based service that is the main goal of this research is very important, especially since this service uses the most up-to-date deep learning techniques to diagnose a person with Covid 19 disease. In this paper, the deep learning approach is used and the Alexent model is used in this process. The data used in this study include 8840 chest images, half of which are related to people without Covid 19 virus involvement and the other half are related to patients with Covid 19 virus. The model proved to be very reliable with 99. 26% accuracy in diagnosis and 95% sensitivity and 99. 7% specificity.

This article aims to design and develop Rapid Prototyping Techniques Ontology based on the study of new generation web as the Semantic Web that is a method of encoding and retrieval of information will be able to understand and process the information. To create an ontology that makes up the backbone of the Semantic Web, first, the selective techniques of rapid prototyping systems were studied, and in the operating area of the appropriate technique, knowledge was extracted with the content analysis method. The output of this process is the ontology of rapid prototyping techniques that are fully covered knowledge in a given area with more than 600 axiom, 120 classes and sub-classes, and more than 60 features. In addition to a knowledge-based view in the field of selector systems of Rapid Prototyping, opens a new arena. In the end, domain knowledge using the owl language in the Proté gé application is implemented as Rapid Prototyping Ontology.

An epidemic caused by a new type of Coronavirus family, called COVID-19, has created a global crisis involving all countries of the world. In this regard, designing an early detection system using heuristic and non-invasive methods can be a good and decisive factor in detecting the disease early and consequently decreasing the prevalence of the virus. In recent years, to rapidly diagnose diseases, machine learning techniques have increasingly grown to predict and diagnose patients, and researchers have used them in various studies. In this regard, since the outbreak of COVID-19, several researchers have tried to use the machine learning approach as a potential tool for identifying and diagnosing this disease. Due to the importance and role of using clinical and laboratory data in the diagnosis of afflicted people with COVID-19, in this paper, the models of K-NN, SVM, Decision Tree, Random Forest, Naive Bayes, Neural Network, and XGBoost as the most common machine learning models were used on a database with 1354 records consisting of clinical and laboratory data of COVID and non-COVID patients to diagnose COVID-19. Evaluation results based on Accuracy, Precision, Recall, and F-Score criteria showed that a XGBoost and K-NN with accuracy of 97% and 96% could be considered a suitable predictive model to diagnose the COVID-19 disease.

One of the algorithms used for sampling complex networks is the classical random walk algorithm, which has been considered due to its good performance. But speed and energy consumption can also be improved by reducing size of input data. In this study, two random walk algorithms inspired by two methods, choosing seed node, and no-retracing algorithm which obtained by changing the classical random walk algorithm, and combining these three algorithms with google page rank algorithm, are discussed. This is done to preserve important nodes and reduce the size of the input data. This sampling was done from the United States flight network database. Also, important characteristics obtained in sampling, such as sampling efficiency, degree distribution, average degree, and average clustering coefficient have been investigated. The algorithms studied in this research each have their own advantages and disadvantages. For example, the no-retracing shows better performance in terms of time and average clustering coefficient. This efficiency is even greater when we use a combination of no-retracing algorithm with google page ranking algorithm. These algorithms can be used when speed is important in decision making, such as deciding on airlines and public transportation, etc. These algorithms are also more energy efficient than the studied algorithms.

In the new generation of industries, Industry 4. 0 plays a prominent role to establish the smart manufacturing. The Industrial Internet of Things (IIoT) is one of the main technologies for realizing the objectives of Industry 4. 0 by enabling the Cyber-Physical Production Systems (CPPS), digital transformation and supply chain. The common communication model in IIoT relies on publish-subscribe model, in which the data collected and stored in a central controller from sensors to be delivered to actuators. This centralized data management is improper for practical IIoT applications on account of its high communication overhead and incompetence for strict delays and energy constraints in IIoT networks. In this paper, to address the aforementioned issues, we propose a Latency-aware and Energy-efficient Data Management Layer (LEDML) to cache data distributedly in some IoT nodes, referred to as proxy nodes, so that the energy consumption and data access latency are jointly optimized while the realistic IIoT constraints in terms of data access delay and cache utilization are fulfilled.

In IoT systems, evaluating the trustworthiness of users and devices before relying on their information is a significant problem. Preserving privacy, decentralization, and self-management without the need of a third party is an open problem in trust management in IoT systems. Collecting reputation feedback and shreds of evidence can also be a source of security issues. On the other hand, blockchain is a promising approach for such environments where centralization and trusted third parties can cause security issues. We present a method by extending Contract Net Protocol for task allocation, which can be exploited as a base for trust management in IoT systems. This method allocates tasks to participants through smart contracts, and feedback values are securely collected in the blockchain. After registration of devices, all communications among IoT devices are handled by smart contracts, and reputation values are maintained in blockchain only when a previous task allocation is successfully taken place. The proposed method is evaluated by experiments on Hyperledger Fabric, and results are reported. Our findings indicate that the proposed method enables the IoT environment to collect feedback values robustly. The technique can be applied in a blockchain-based trust management model for security enhancement.

Review spam is an opinion written to promote or demote a product or brand on websites and other internet services by some users. Since it is not easy for humans to recognize these types of opinions, a model needs to be provided to detect these types of opinions. In recent years, a lot of research has been done to detect these types of reviews, and with the expansion of deep neural networks and the efficiency of these networks in various issues, in recent years, various types of deep neural networks have been used to identify spam reviews. This paper reviews the proposed deep learning methods for the problem of review spam detection. Challenges in this area, evaluation criteria, and datasets in this area are also reviewed.

Almost all security protocols of wireless sensor networks believe that the enemy or attacker can take full control of a sensor node through direct connection. Security is very important in accepting and using sensor networks in many applications. In order to clarify this issue, we focus on detecting anomalies in the nodes and cluster heads of wireless sensor networks, and look for a solution to detect anomalies in the nodes and cluster heads and determine new cluster heads. A group of researchers to detect anomalies have suggested Mobile Data Collectors (MDCs) machines, where some abnormal nodes may be inactive at the time of inspection and not be identified, and due to environmental problems, the machine cannot go to those places, it is also very expensive and cannot work online and cannot quickly overcome attacks. Due to the large number of sensors, it is not scalable. In this article, we first review the methods that have been proposed until now and describe their advantages and disadvantages and then propose a method that detects the anomalies of the nodes in the cluster heads and detects the anomalies of the cluster heads in the sink node, it runs without the need for external circuits and does not impose additional costs, it works online and can quickly overcome attacks. Our proposed method for evaluating performance was simulated by MATLAB software and it uses Intel Research Laboratory Database.

Recent studies have been indicating that many clinical drug combinations surpass single-drug therapy efficacy. Machine learning, deep learning, network analysis, and search algorithms have been considered to facilitate the discovery of synergistic drug combinations, and two of the best state-of-the-art models in this area are under the deep learning category. In this paper, we present DComG, a Graph Auto Encoder method to predict synergistic drug combinations. Using the dataset provided in DCDB, our analysis shows tremendous improvement in the performance of predicting new drug combinations over previously introduced state-of-the-art models by an average of 4% in ROC_AUC scores. We highlight the importance of drug-drug interactions (DDI) in the form of node2vec features of DComG graph inputs for predicting new drug combinations. Finally, we address the results of our model in terms of biological interpretations of drug combinations based on recent medical drug combination papers in the literature.

This study was conducted to determine the mixed-method usability evaluation of the Iranian national covid-19 electronic screening system. The cross-sectional study was carried out in partnership with 116 users of the Iranian national covid-19 electronic screening system and five experts. As a result of the experts' assessment, the Iranian national covid-19 electronic screening system scored 0-2 out of the 10 principles of Nielsen Jacob, which indicates a good approach to the design of this system. To evaluate, the questionnaire for user interaction satisfaction (QUIS) version 7 was used. Data were analyzed by spss version 19. A total of 112 out of 116 questionnaires were obtained. In the Iranian national covid-19 electronic screening system, nine (33. 3%) of the 27 sections scored higher than seven. More than half scored over five. There were no factors in the terminology and system information and learning section between 7 and 9; the highest rankings in the section overall responses to the software were 1) terrible-wonderful 2) difficult-easy; in the section overall reactions to the software, all of the factors were highest; also, the highest rankings were in the section "system capability" for 1) system speed 2) system reliability 3) designed for all levels of users.

It is nearly three year since lunching the transparency website in the Tehran municipality and other big cities of Iran with the purpose of fighting against financial and administrative corruption there. The transparency via open data approach, which spreads by new communication and information technology, basically shows that it can be considered as a media phenomenon. Now it is time to estimate its impacts on the cyberspace in Iran to show that how much this system has approached its purposes and where it is heading to. The main focus of the research is on the municipality of Tehran as the biggest city and the capital of Iran. Recognizing websites linked to the Tehran municipality transparency system, as representatives of social groups, can help us to survey the impacts of this system on Iran society much better. This article was looking for the number of websites linked to the online Tehran municipality transparency system by linkage analysis research method with the help of Webometric software and after exporting these websites statistically, they were semantically grouped to better understanding of their social characteristic and intended purposes. Most of the referred and linked websites to the Tehran municipality transparency system were the report and political sites, which shows the political importance of transparency movement in Iran. It seems that this movement is getting political. It can be in conflict with the financial and administrative nature of transparency, which is a specialized matter.

With growing use of internet and online applications, network traffic classification could be much more useful nowadays, because managing network services and quality assurance, two key points in network structure, could be done easily using this kind of classification. Different methods are used for this task, including port-based classification, machine learning and some other algorithms that each of them had its own advantages and disadvantages. For eliminating such disadvantages, deep learning methods are new ways for doing this task due to the power and excellent performance they showed. Furthermore, most of the work done in this field are using non-encrypted traffic or encrypted traffic in mobile networks, but as we know, privacy of data is very important these days. In this article, with the use of deep learning neural network, encrypted traffic of non-mobile data is being classified. For this purpose, we use the UNB ISCX VPN-non-VPN dataset that includes encrypted and unencrypted traffic of different applications. Then we design an algorithm based on DNN that could classify these traffics effectively. Performance of the model was evaluated and 0. 86 accuracy and 0. 78 f1-score showed that model works well compared to other algorithms used in this area.

The Internet of Things can be considered as a set of intelligent devices, which interact to achieve one or more specific goals. Smart devices use Internet protocols to communicate with other entities. One of the challenges of communication between nodes is trusted routing to send packets. In this paper, a trust management algorithm with a trust rank-based clustering method is proposed to improve IoT security that can detect malicious nodes and malicious connections and send packets from the trusted route to the destination in the fastest time. The purpose of this algorithm is to deal with fake nodes and connections. The simulations were performed in MATLAB environment and the proposed method was compared with one of the known methods called Self-Channel Observation Trust and REputation System(SCOTRES). The simulation results show that the proposed method has better performance in insecure communication, average lost packets and average trust and average received packets than the base method.

Aparat has become one of the largest Iranian websites on the Internet which also has induced Social Network features. This article provides the factors affecting the popularity of social networks and uses this information based on Aparat’ s attributes. A content analysis of 618 videos from 13 Aparat channels was conducted. The main purpose is to find solutions or formulate that change each influence parameter to quantity knowledge and then make the connection between these parameters to get a final score for each channel. This number is influenced by all the important factors affecting the popularity of the channel and that is the reason for its impotence. Moreover, it is used to rank channels based on popularity.

Incoming images are always exposed to noise. A clear example of noise is salt pepper noise that enters a gray image due to channel disturbance. This article presents a new method for removing salt pepper noise using Riesz filters. First, the basic concepts are explained and the Riesz average parameter is introduced. Then, in order to reduce the noise effect, an algorithm based on the mean of the Riesz is proposed, which has a high ability to remove salt pepper noise. The proposed algorithm is applied to several gray images of the sample and the results are presented quantitatively and qualitatively. The two criteria of structural similarity and maximum signal-to-noise ratio are considered in this paper, and the results are compared with the four conventional methods recently reported in this field. The simulation results indicate that the proposed algorithm achieves higher indicators in terms of quantity than other proposed algorithms in this field and has a better performance in terms of quality than noise removal.

There is 10% fraud in medical insurance based on published statistics in Insurance Research Institute of Islamic Republic of Iran in 1399 – solar system eq. 2020 in the Gregorian calendar-which cost about 28 thousand billion RIALs – the official currency of Iran eq. to about 320 million dollars-. This study proposes a machine learning-based technique to predict the claim cost based on other patients' history and predict fraud or abnormal costs in claims that significantly differ from other claims. Besides, a new data sampling approach is proposed to lead the machine learning algorithms that focus on exceptional cases. A real-world private dataset is used to evaluate 700, 000 claims of the RASA web portal, used for supplementary insurance by famous companies like Day. The proposed data sampling approach reduced absolute error in exceptional cases from 35 to 23 errors for deduction rate. The evaluation results show about 0. 5% of abnormal cases in the dataset with a higher than 20% absolute error. The abnormal rates can be adjusted to a lower or higher range.

The mass volume of information posted on social networks has attracted increasing attention due to the variety of applications that can be used. In fact, these virtual communities enable users to express their viewpoints as well as influence other users. Therefore, it is very important to identify the key influencers of the social network and the communities formed around a topic, using the method of social network analysis. The connection between different entities in the network will help to understand this issue. In this study, conversations about the new coronavirus have been examined on Twitter. The illustrated network is examined in Gephi software by one of the community detection algorithms such as Louvain. HITS and PageRank algorithms are also used to analyze the impact of information propagation and ranking of entities.

Blockchain networks are already extensively used in various applications because of their increased security. The unique characteristics of blockchain technology, such as decentralized, peer-to-peer, and invariable distributed ledger qualities, make it appealing to researchers, academics, and industry. The consensus protocol is a fundamental part of blockchain technology. PoW (Proof of Work) or fixed-validator consensus protocols comprise most of the existing consensus mechanisms. However, the tremendous computational effort required for PoW leads to excessive energy and computing resource usage. On the other hand, Fixed-validator protocols validate new blocks by a fixed, static set of validators, allowing attackers to execute multiple attacks against these validators. In this article, we proposed a novel consensus protocol base on the Proof of Activity protocol and game theory. Our consensus protocol is efficient in energy consumption and can deal with selfish mining and majority-attack.

This paper describes the creation of Exa Persian Paraphrase Corpus (ExaPPC), a large paraphrase corpus consisting of monolingual sentence-level paraphrases using different sources. ExaPPC is the first large-scale paraphrase dataset used in Persian paraphrase detection to the best of our knowledge. There are 2. 3M labeled sentence pairs in the corpus consisting of a 1M paraphrase label and 1. 3M non-paraphrase label. Efforts were made manually and semi-automatically to construct this corpus using techniques such as subtitle alignment, translating existing parallel English-Persian corpus and similarity corpus on English tweets. In addition to enriching the corpus, candidate sentence pairs among tweets have been extracted via NLP tools and labeled by two Persian native speakers. The advantages of this corpus compared to the existing ones are the number of pair sentences, sentence Length variation and textual diversity, including formal and dialogue sentences. The result on the provided test corpus shows that ExaPPC achieves 94% accuracy on paraphrase detection task. The corpus is publicly available1.

Crime is a behavioral disorder with various scales that are intimately linked to a variety of circumstances such as spatial, temporal, sociological, and ecological aspects. The massive amounts of crime-related data, which is being published and grows with each passing day, in the form of online news reports have prompted researchers to pursue studies in the field of violence and criminal investigations. In this work, we developed a semantic approach to extract spatiotemporal and crime-related information from news reports to detect crime spatial distribution. The proposed method, in particular, aims to extract geographical and temporal information to detect regions with a high number of criminal cases, as well as to represent semantic knowledge of criminal incidents by annotating spatiotemporal information from their web domains. This approach incorporates the use of Natural Language Processing (NLP) techniques and a crime domain ontology into the information extraction process to automatically retrieve spatial, temporal, and other relevant information about criminal behavior from news reports. Our proposal consists of a comprehensive solution built on a fully functional architecture that has been tested in a use case scenario for the crime news reported in London, United Kingdom.

With the rapid growth of computer devices, network communication faced different challenges from network management to traffic engineering. Software-Defined Networking (SDN) is a well-known solution for optimizing these communications. SDN is a new networking architecture to simplify network management that separates the control plane from the data plane. The central controller is the major advantage of SDN; however, it has security vulnerabilities such as being unreachable in Distributed Denial-of-Service attacks (DDoS). Consequently, it is very important to protect SDN from DDoS attacks. In this paper, we proposed an algorithm for DDoS attack detection and reducing its impact in SDN architecture with multiple distributed controllers. We presented two methods 1) the entropy of destination IP addresses and 2) Packet window initiation rate for early detection of DDoS. We used Mininet and floodlight to simulate our algorithm in different scenarios. The result shows that our algorithm outperforms other works in various network configurations and multi-victim attacks.

Today, with the development of the world of communication and the emergence of social networks, people's lives and decisions have undergone many changes. Social networks with the aim of informing, have been able to play a major role in various aspects of life and decision making. In the meantime, people use social networks to make various decisions such as investment and type of stock. In this paper, a method based on deep learning is designed and developed to predict stock value movement. The data was collected from the social networks Twitter and Yahoo Finance for three months. Twitter data is first tagged with semantic analysis and then tagged using CNN. Stock market value data was calculated along with the emotion index and analyzed by different modes of the proposed LSTM model to predict the trend of stock market value. The results show that emotion indices and calculation of HLPCT and PCTchange criteria have been effective in predicting the trend of stock market value with the least error. The results also show the superiority of the proposed method compared to the existing methods in terms of price prediction.

Nowadays, considering the huge amount of DATA, to search through them, we ought to use methods for analyzing the DATA according to our needs. This challenge also exists in the entertainment and cinema industry to find movies and TV shows with the same topic aiming to recommend and minimize the search space for the audience. Therefore, methods are needed to efficiently recognize the movies with the same topic and present them to the users. Most of the existing services lean on user-based information, and usually, not on the original content of the movies. These services use DATA such as user ratings and comments or features like actors, directors, and the movie genre or a combination of both. In this paper, we use low-level features of the movie subtitles, extracted using LDA, for thematic analysis of textual contents of the movies (subtitles). To do so, using the extracted features and Cosine similarity measure, we construct the similarity graph of movies. In this graph, each node represents a movie and each edge indicates the similarity between them. In the following, using clustering methods on movies graphs we were able to achieve a noticeable Thematic correlation between the movies.

The study of social networks analyzes the relationships between humans, organizations, and interactions between entities. The wide-spreading popularity of social media has attracted researchers' attention on community detection. Communities are defined as clusters of nodes or vertices that have stronger relationships between entities inside a cluster than relationships between clusters. Community detection plays an important role in discovering the underlying structures of social networks and it displays the effects of links' structures on people and the relationship between them. Usually, clustering algorithms are utilized to identify the communities. In this paper, we propose a new clustering algorithm for community detection that considers the depth of relationships between individuals in the community identification process. Results on two popular datasets indicate that considering the depth of relationships improves the accuracy of the clustering methods. From the compared results, the proposed algorithm outperformed the six state-of-the-art algorithms.

Social network analysis with large volumes of data and complex communication structures is so difficult and time-consuming. Community detection is one of the major challenges in network analysis. A community is a set of individuals or organizations whose communication density is more than other network entities. Community detection or clustering can reveal the structure of groups in social networks, or relationships between entities. The label propagation algorithms with neighbor node influence have less complexity than traditional algorithms, such as clustering, to recognize communities. Also, the algorithms can identify overlapping communities. In our label propagation algorithm, which is based on the neighbor node influence, important nodes are more likely to publish their labels, while less important nodes have a small chance of spreading the label. The degree of similarity of nodes and the effect of nodes in a social network depends on the parameter of path length between nodes. In the proposed method, increasing this parameter leads to more accurate identification of overlapping and stable communities. The proposed algorithm detects overlapping communities with the same accuracy as the previous algorithms with fewer iterations, in less time. The algorithm is implemented on real and artificial social networks with weightless graphs and weighted graphs with weighting by Jacquard similarity criterion, in all of which the execution time is improved.

With the proliferation of smartphones, smart solutions to various problems have been provided. Today, due to the large volume of out-of-home advertising, its impact on the audience has decreased. Smartphones are a suitable and new tool to display and increase the effectiveness of this type of advertising because of their power of customization. Out-of-home advertisements are displayed instantly on the user's mobile phone when being in indoor places such as stores. One of the challenges is to calculate the exact location of the user in indoor places due to improper functionality of GPS signals. In this article, an indoor location-based solution using BLE beacon signals is proposed. The proposed method includes decreasing signal fluctuation, send and receive data simultaneously through beacon devices equipped with low-power Bluetooth (BLE) technology, and finally build an efficient model to predict the user location using decision tree algorithms. According to the implementation results, the indoor positioning is achieved with an accuracy of 1 meter and 13 cm within a specified range and with only 8 beacons, and in 98% of cases, the predicted point is correct.

Spam is an example of unwanted content sent by unknown users and causing problems for mobile phone users. Disadvantages of spam include the inconvenience to the user, the loss of network traffic, the imposition of a calculation fee, the occupation of the physical space of the mobile phone, the misuse and fraud of the recipient. For this reason, the automatic detection of annoying text messages can be fundamental. Also, recognizing intelligently generated text messages is a challenge. Nevertheless, the current methods in this field face obstacles, such as the lack of appropriate Persian datasets. Experiences have shown that approaches based on deep and combined learning have better results in uncovering the annoying text messages. Accordingly, this study has attempted to provide an efficient method for detecting SMS spam by integrating machine learning classification algorithms and deep learning models. In the proposed method, after performing preprocessing on our collected dataset, two convolutional neural network layers and one LSTM layer and a fully connected layer are applied to extract the features are applied on the data which forms the deep learning part of the proposed method. The Support vector machine then utilizes the extracted information and features to perform the final classification, which is a part of the Machine Learning methods. The results show that the proposed model implements better than other algorithms and 97. 7% accuracy was achieved.

Accessing people's personality traits has always been a challenging task. On the other hand, acquiring personality traits based on behavioral data is one of the growing interests of human beings. Numerous researches showed that people spend a large amount of time on social networks and show behaviors that create some personality patterns in cyberspace. One of these social networks that have been widely welcomed in some countries, including Iran, is Telegram. The basis of this research is automatically identifying users' personalities based on their behavior on Telegram. For this purpose, messages from Telegram group users are collected, and then the personality traits of each member according to the famous NEO Personality Inventory (NEO PI-R) are identified. For personality analysis, the study is employed three methods, including; Cosine Similarity, Bayes, and MLP algorithms.

Every insurance company plays various roles and so implements many processes. One of the essential processes is the assessment which occurs due to investigating bills and claims and may lead to cost deduction based on fraud or limitation rules. This process may go forward in one or more steps by one or more assessors; some last some minutes, and others last more. This study tries to investigate actions, durations, assessors, and everything related to this process to help the insurer companies know useless steps, verify assessor's behaviors, and make better decisions for the assessment process. This paper aims to demonstrate the methods and efforts of each mining rule and its strengths and weaknesses. The clustering and association rules mining are proposed for this objective. The gathered dataset has about 110, 000 records from the Rasa web portal and is selected from an Iranian insurer company's one-year indirect assessment process. The results show that some steps are useless, some assessors do not assess at an appropriate duration, and some bills consider several times with one assessor, which is suspicious behavior.

Community Question Answering has a high place in the lives of all members of a society nowadays. It is important for the owners of a community to be able to make this community better and more reliable. One way to achieve this is to find users who have more knowledge, expertise, experience and skills in the field and can well share their knowledge with others (which we call experts and aim to encourage them to be more active in the system). One method to be used is to identify expert users, and whenever a new question is asked, we suggest this question to them to check and answer if its in their scope of expertise. One way to encourage users to post replies is to use gameplay techniques such as assigning points and badges to users. But as we will discuss, this method does not always reflect expert users well, because some users will try to have small and insignificant but numerous activities that will make them gain a lot of points, however they are not experts. In this study, we examine the methods by which experts in a question-and-answer system can be found, and try to evaluate and compare these methods, use their ideas and positive points, and add our own new ideas to a new way of finding them.

The purpose of this study is to study the impact of Information and Knowledge Science Faculty Members and science of ten first level universities in Iran in the semantic scholar search engine. The present study is a survey and the scientometric approach has been used. The statistical population of the study was 53 faculty members identified in this search engine. Both descriptive (Excel software) and inferential (SPSS software) statistics utilized to analyze data. The results showed that out of the total citations received by the research outputs of the faculty members, 5% of citations are Highly Influential Citations and Ali Norouzi is the most influential researcher. citation to effective research output are often in the introductory sections of papers and is done to provide the facts, definitions, concepts and basic information. To be more effective, researchers need to be actively involved in this search engine and use its capabilities to be visibility.

The current trend in the field of computer systems is moving in the direction that users, instead of launching the services they need on traditional data centers, procure them from providers that provide these services through the Internet and are responsible for setting them up and maintaining them. These centers have a pool of resources that create the image of unlimited resources for users. Also, the services of these providers benefit from a high degree of flexibility, as a result of which the client can receive exactly the resources corresponding to their needs and pay the same amount. With the expansion of cloud data centers, the placement of virtual machines in existing systems with awareness of factors such as security, energy and data traffic between different machines has become main concerns of these centers. This is because not paying attention to these factors and using random, greedy or intermittent algorithms entails high financial and security costs for cloud service providers. In this study, we were able to provide a way to improve the security and location of virtual machines compared to other methods using a more accurate and appropriate decision to improve each, respectively, by an average of 15% and 23%.

The present study was conducted with the aim of technical and legal Pre-feasibility study of using social networks in Iranian public libraries. In this research, Telos method has been used to pre-feasibility of using social networks in Iranian public libraries, focusing on its technical and legal field. In the first step, the pre-feasibility document format was developed based on library studies as well as two sessions of the nominal group. In the second step, based on the researcher's triangulation, the initial pre-feasibility document was developed. Finally, in the third step, in order to validate the initial pre-feasibility document obtained, a survey of experts was considered. In this step, semi-structured interviews were conducted with nine experts in this field, who were selected by Targeted sampling; and basic document of the Pre-feasibility study of using social networks in Iranian public libraries was finalized. Based on the findings of this research, the document of technical and legal Pre-feasibility study of using social networks in Iranian public libraries includes sections: 1-introduction, advantages and limitations, 2-scenarios, 3-technical considerations and 4-Legal considerations. The details of these sections provide a codified set of technical and legal information on the implementation of social networks to the managers of the Iranian public libraries.

The Internet of Things (IoT) has introduced new advances in sensors, Internet technologies, and communication protocols. Sensors in the IoT network have the important task of collecting information from the environment in which they are located. These sensors face important limitations such as limited energy resources, data storage, transmission, and processing power. In addition to the inherent limitations of the network's sensor nodes, they are vulnerable to a variety of security threats due to the presence of invasive and malicious nodes. In previous research, less attention has been paid to the issue of energy conservation and secure routing. Accordingly, in this paper, an energy-aware and secure protocol for routing data in the Internet of Things is presented. This protocol detects and prevents data compromise by simultaneously using the heuristic and Elliptic-curve Diffie– Hellman (ECDH) cryptographic method. The evaluation shows the proper performance of the proposed protocol compared to competing protocols. The results showed that the protocol managed to improve the metrics for the number of lost packets by 13. 44%, the throughput by 86. 55%, and the power consumption by 0. 022% for 250 nodes.

Nowadays, due to the important role of software systems in our lives, without the use of software, we will not be able to get most of the services we are used to. The quality of these services depends on the quality of the software that implements them. There are many criteria for measuring the quality of a software. The presence or absence of antipatterns can be a measure of software quality. Some well-known Antipatterns are studied according to their effect on performance, reliability and other related criteria. One of these criteria is the readability of the software source code. Developers of a program are not necessarily the only ones developing it in the future, so it is important to follow tips that make it easier for potential developers to understand how the program works. This article examines the linguistic antipatterns associated with naming functions and provides solutions for their automatic resolution. Antipatterns that are related to the naming of members of the software source code are called linguistic antipatterns. This paper presents a method for automatic detection and elimination of these antipatterns using abstract syntax tree. The proposed method is then tested on the source code of several open source softwares.

Choosing the optimal software architecture in the search space by considering quality criteria is beyond human capabilities and is very challenging. It is necessary to search the design space automatically to improve the existing architectural features. To do this, we can use search-based software engineering approaches. In this study, we examine the methods of optimizing and evaluating software architecture and provide a search-based method to improve the reliability of software architecture. The proposed method is based on the use of NSGA-II algorithm and genetic programming and the use of software architecture reliability tactics in it. In the proposed method, we optimize the software architecture in two steps. First, we use the genetic programming algorithm to extract how to apply the software architecture reliability tactics, and in the next step, we use the NSGA_II algorithm to search for the optimal allocation of components to the hardware servers. To evaluate the proposed method, we use a reporting system case study. The results of applying the proposed optimization steps show that the reliability of the whole system as well as most of its most frequent functionalities is improved.