mohammadjafar tarokh

To develop a deep learning model to predict the risk of cardiovascular diseases (CVDs) events based on features found in fundus images.

We developed a predicting model for cardiovascular diseases based on retinal fundus images using the deep learning method. We trained our model using 2,091 retinal fundus images obtained from 211 patients. Our dataset included demographic information of each person, conventional CVD risk factors, CVD risk estimated number (calculated using the Framingham method), strokes and heart attack incidents during 5 years (patients who were referred to the ICU or CCU), and retinal fundus images for each person. We used receiver operating characteristic (ROC) analysis to assess the accuracy of our classification model.

Our proposed algorithm was able to identify high-risk individuals from no-risk individuals with 83% accuracy and a high confidence level (AUC = 0.91, P value < 0.0001). The results also showed that our model could predict cardiovascular events such as stroke with a probability of 72% (AUC = 0.83, P value< 0.0001). In comparing our model’s ability to predict CVD risk with the Framingham risk score, the Framingham model’s accuracy was 65 % in our dataset (with a best AUC of 0.78).

Our deep learning prediction model developed based on retinal fundus image findings to predict the risk of CVD, showed a relatively high accuracy. Its accuracy was higher than traditional prediction models like the Framingham model and comparable to other models based on fundus images for predicting CVD.

In recent years, machine learning and evolutionary algorithms have drawn the attention of researchers and specialists in various fields, especially in healthcare, due to their practical applications in processing large datasets to provide valuable insights. Considering the increasing prevalence of diabetes and its rapid and accurate diagnosis being one of the most critical issues in medicine, significant concerns are faced by global communities worldwide. The present study was conducted with the aim of creating a diagnostic model based on evolutionary algorithms and machine learning to diagnose diabetes.

This research based on the Indian Pima diabetes dataset presents a framework based on intelligent diabetes diagnosis. The proposed method consists of two main stages. The first stage involves a classification approach using K-nearest neighbors and random forest algorithms. The second stage includes a combined feature selection and classification approach to enhance the results of the first stage, utilizing grey wolf optimization, whale optimization, and particle swarm optimization algorithms for feature selection. Comparative analysis among different approaches is conducted through evaluation metrics such as accuracy, precision, recall, and F1-score.

After comparative comparisons among the proposed models, the random forest model based on the grey wolf optimization was selected and introduced as the final model with a prediction accuracy of 81.38%.

The findings of this research indicate that the use of evolutionary algorithms alongside machine learning models can often enhance the efficiency and accuracy of diabetes diagnosis and its associated complications.

Customer churn presents a significant challenge for the telecommunications industry, necessitating effective strategies for prediction and prevention. While prior research has explored diverse methodologies, including Agent-based Modeling (ABM), limitations persist. Existing approaches often rely heavily on theoretical constructs, resulting in oversimplified models and constrained data utilization. This study addresses a critical research gap: the absence of a comprehensive framework integrating empirical data, agent-based modeling, and machine learning techniques for churn prediction in telecommunication markets. By bridging the gap between theoretical abstraction and empirical reality, proposed framework enables more proactive churn management strategies. Additionally, it facilitates the simulation of diverse market scenarios, empowering stakeholders to optimize key metrics such as revenue and market share. Through the implementation of the proposed framework within a specific telecom market scenario involving two competing entities, this study demonstrates its efficacy in achieving desired market share objectives. This research contributes to advancing the understanding and management of customer churn in the telecommunications industry.

Pulmonary emphysema is one of the lung diseases that usually remains unknown until old age and does not have a definitive treatment. A quick diagnosis of this disease helps a lot to the people involved in this disease and prevents the growth of emphysema masses. This research tries to in early diagnosis of this disease with the help of deep learning methods.

This research tries to diagnose this disease faster with the help of Unet neural network optimized with GPC meta-heuristic algorithm. The data of this research were collected from Imam Ali and Bu Ali Sina hospitals, Zahedan city, Sistan and Baluchistan province. The data include 300 pieces with emphysema, including 65 cases of CLE, 97 cases of PSE, 138 cases of PLE, and 45 cases of normal data. These data were analyzed by Unet deep neural network and GPC optimization algorithm, and finally, with the help of accuracy criteria, recall, specificity, and F-measure were compared and investigated with other methods.

In this research, the criteria used have much better results compared to other emphysema diagnosis methods with the help of the optimized Unet network, with accuracy of 18.97, prediction of 40.98, sensitivity of 48.23, and f score of 97.50, respectively, which shows a faster, more accurate, and more effective diagnosis of this disease with the help of the proposed method.

Using the right deep learning methods in combination with strong optimization algorithms can enable faster and more accurate treatment of diseases.

This paper discusses the integration of Quality Function Deployment (QFD) and Data Envelopment Analysis (DEA) to prioritize technical specifications based on customer expectations and market competitiveness. It aims to enhance product/service quality by utilizing QFD to identify key technical requirements and DEA to evaluate the relative importance of these requirements, ultimately improving customer satisfaction and market competitiveness.

We utilize the Analytic Hierarchy Process (AHP) and DEA methodologies to evaluate technical specifications prioritization in QFD and employs the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method to assess the prioritization of technical requirements based on relative importance weights and limited resources.

The integration of Fuzzy Analytic Hierarchy Process (FAHP) and DEA determines the relative importance weights of technical requirements and their impact on market competitiveness.

Originality/Value: 

The proposed methodology identifies key technical requirements using QFD and evaluate their relative importance using DEA that leads to improving customer satisfaction and market competitiveness of Rafsanjan market tile making unit.

When, we consider the ubiquity of Facebook, twitter, LinkedIn, it is easy to understand how social media is woven into the fabric of our day-to-day activities. It is a suitable tool to find information about news , events , and different Issues. After corona virus outbreak, it is inspired users to understand pandemic news, mortality statistics and vaccination news. According to evidence, the diffusion of pandemic news on social medium has increased from 2020 and user face a ton of COVID19 messages. The purpose of this paper is to determine the check-worthiness of news about COVID-19 to identify and priorities news that need fact-checking. We proposed a method that is called CVMD. We extracted the feature of content. We use the deep learning approach for prediction it means that we model this problem with a binary classification problem. Our proposed method is evaluated by different measures on twitter dataset and the results show that CVMD method has a high accuracy in prediction rather than other methods.

One of the complicated processes in the Ministry of Health is the process of registering pharmaceuticals .Since there are different soft wares for registration of products, and since parallel services concuss waste of time and money, it is necessary to provide integrated software in the form of services. This goal can be achieved when services are identified and combined in integrated soft wares.

present study is an applied research, In order to identify the services of an integrated system for recording pharmaceutical supplies, first the Gray-Wolf multi-objective optimization (GWO) algorithm was proposed. Then the values of the algorithm parameters were extracted by the goal-based requirements analysis method and the algorithm was implemented. Finally the best services were extracted by the hierarchical analysis process.

Considering that the results of present study are an operational project in the Ministry of Health, by implementing the algorithm proposed by Gray Wolf, services were identified, which can be used to create integrated software for registering pharmaceutical supplies.

Creating an integrated system for registering pharmaceutical supplies is one of the important challenges of the Ministry of Health. This can be achieved by identifying services and combining these services to create an integrated system.

Nowadays, the stock market plays an important role in the economy of different countries. The abundance of data in the stock market and the need for fast and correct data processing and making appropriate decisions have made the use of computers inevitable. In this article, using deep reinforcement learning, a model is designed to present the duties of a trader in the Iranian stock market with regard to liquidated shares. In the first step, the history of stock prices along with the indicators based on it are given as inputs to the convolutional neural network. In the next step, in order to calculate the matching rate of the convolution output with the expected output, the sum of squared error cost function is used, which, in turn, is minimized in the optimization process. Since the data in the Iranian stock market is limited, using the convolution model instead of the Q table in the deep reinforcement model prevents the over fitting of the model. In order to evaluate the model, the data of Tehran Stock Exchange was used in the period of 1390 to 1400. The performance of the proposed method was compared with the buy and hold strategy. The results show, in some cases, the profit from the proposed method in contrast with buy and hold strategy is 21% and -7%, respectively.

The presence of social media has brought about fundamental changes in today's societies. These media are a good platform for users to share ideas. This has caused users to encounter a lot of information that is not suitable for them most of the time and has little influence on them. Providing a way to select effective posts for the user among a multitude of posts can be very important. The methods presented in recent research for selecting effective posts are based on statistical features related to various microblog data and less in terms of content content, each post has been measured on a specific user. Despite the variety of topics, content of tweets and different users, most of these methods are not accurate by providing a general model based on a large number of features and are not able to provide online predictions. In this study, by analyzing the dissemination of posts among users in a specific period of time, a method for measuring users' attention to shared content and their effects is examined. This method is named IKS, which is based on the characteristics of content published by The user is presented as a binary classroom problem based on in-depth learning. Evaluation of this method has been done using intuitive method and data set evaluation which is more accurate compared to other methods.

One of the complex processes in the Ministry of Health and Medical Education in Iran is the process of registering pharmaceutical supplies. Currently the registration process is a multi-stage process, resulting in parallel services, a waste of time and unnecessary expenses. Therefore, an integrated system will improve the relevant service delivery. The purpose of this study was to identify a set of activities that could be collectively considered as a single service to create an integrated system for registering pharmaceutical supplies in the Ministry of Health and Medical Education, Iran.

This was an applied research, beginning with collecting information about the registration processes for 20 different products. In order to identify the services/steps of an integrated system for recording pharmaceutical supplies, first the Gray-Wolf multi-objective optimization (GWO) algorithm was proposed. Then the values of the algorithm parameters were extracted by the goal-based requirements analysis method and the algorithm was implemented. Finally the best services were extracted by the hierarchical analysis process.

Through the proposed algorithm seven services were identified, including product class inquiry, document registration, document review according to license type and product class, licensing, laboratory services, clinical studies and payment service. These services were carefully approved with a precision of 97.3 by the experts of the Ministry of Health. The proposed framework for recording drug requirements was found to be effective and could facilitate the process by up to 90%, reduce the processing time by 80% and reduce the processing costs by 65%.

Creating an integrated system for registering pharmaceutical supplies is one of the important challenges of the Ministry of Health and Medical Education. This can be achieved by identifying services and combining these services to create an integrated system.

Fuzzy expert systems are intelligent systems which can be used to obtain better results in evaluating the performance of the banking system. The purpose of this study is to evaluate the performance of bank branches using fuzzy variables beside financial variables. In this study, firstly, the rules of the data were extracted by implementing data mining algorithms on the financial data of branches. In the next step, by obtained rules of financial data and along with fuzzy variables, a fuzzy expert system is designed in order to achieve a system that can comprehensively evaluate the bank branches performance. For designing the considered expert system, nine fuzzy variables such as branch location, customer loyalty, employee satisfaction, customer satisfaction, creativity and innovation, branch appearance, staff appearance, employee stability and also the output of financial rates have been used. Decision tree and C.5 algorithms have been used in order to extract the rules in the branch data. MATLAB fuzzy inference system has been used to design the fuzzy expert system also. The results of the research illustrated the hidden knowledge of the branch data can be extracted via data mining and the performance of bank branches can be evaluated as a comprehensive information system by fuzzy expert systems.

Branches of Bank are one of the most important pillars of digital banking and surveying their performance plays an important role in profitability and achieving the bank's goals. This study evaluates the performance of bank branches using innovative methods. First, important indicators for evaluating branch performance have been identified. Then, the proposed method for data of bank branches has been implemented in the form of a case study. For this purpose, clustering was done first to separate efficient, semi-efficient and inefficient branches. Then, based on the labels created on the data of the branches, classification algorithms and decision trees were used to extract the rules in the data of efficient, inefficient and semi-efficient branches. In the present study, the proposed model of C5.0 algorithm was used due to obtaining the highest accuracy compared to other algorithms. Finally, based on the extracted rules, an expert system was designed to evaluate the performance of bank branches. Clips software was used to design the expert system. In the bank under study, the average increase percentage of cheap deposits during the period to increase the target balance had the greatest impact on performance.

Alzheimer chr('39')s disease is the most common form of dementia which has caused disorder in memory. Cerebral palsy and posttraumatic stress disorder (veterans of war, warriors, armed forces) play an important role in increasing the risk of Alzheimerchr('39')s disease. The nature of the large dimensions of neural data, as well as the small number of available samples, make for an accurate computer diagnostic system. The aim of this study was to apply deep neural networks to develop an automatic disease diagnosis system.

Methods and Materials: 

In this research, studies on magnetic resonance imaging of war veterans are done by python Software. In the proposed model, in the proposed model, 10% of the images of the data base were selected for training. In the first stage, the training is from deep learning with Convolutional network to extract the features, then in the second stage, in order to classify the health status based on the learned features.

The results of the analysis are also compared with the results presented in previous Studies. The proposed method has higher detection accuracy than the existing Ones, which increases the accuracy of detection in many cases.

The results of this study showed that using intelligent methods based on deep learning can accurately diagnose the disease.

 Currently, diabetes is one of the leading causes of death in the world. According to several factors diagnosis of this disease is complex and prone to human error. This study aimed to analyze the risk of having diabetes based on laboratory information, life style and, family history with the help of machine learning algorithms. When the model is trained properly, people can examine their risk of having diabetes.

 To classify patients, by using Python, eight different machine learning algorithms (Logistic Regression, Nearest Neighbor, Decision Tree, Random Forest, Support Vector Machine, Naive Bayesian, Neural Network and Gradient Boosting) were analysed. were evaluated by accuracy, sensitivity, specificity and ROC curve parameters.

 The model based on the gradient boosting algorithm showed the best performance with a prediction accuracy of %95.50.

 In the future, this model can be used for diagnosis diabete. The basis of this study is to do more research and develop models such as other learning machine algorithms.

Automatic semantic segmentation of skin lesions is one of the most important medical requirements in the diagnosis and treatment of skin cancer, and scientists always try to achieve more accurate lesion segmentation systems. Developing an accurate model for lesion segmentation helps in timely diagnosis and appropriate treatment.

In this study, a two-stage deep learning-based method is presented for accurate segmentation of skin lesions. At the first stage, detection stage, an approximate location of the lesion in a dermoscopy is estimated using deep Yolo v2 network. A sub-image is cropped from the input dermoscopy by considering a margin around the estimated lesion bounding box and then resized to a predetermined normal size. DeepLab convolutional neural network is used at the second stage, segmentation stage, to extract the exact lesion area from the normalized image.

A standard and well-known dataset of dermoscopic images, (ISBI) 2017 dataset, is used to evaluate the proposed method and compare it with the state-of-the-art methods. Our method achieved Jaccard value of 79.05%, which is 2.55% higher than the Jaccard of the winner of the ISIC 2017 challenge.

Experiments demonstrated that the proposed two-stage CNN-based lesion segmentation method outperformed other state-of-the-art methods on the well-known ISIB2017 dataset. High accuracy in detection stage is of most important. Using the detection stage based on Yolov2 before segmentation stage, DeepLab3+ structure with appropriate backbone network, data augmentation, and additional modes of input images are the main reasons of the significant improvement.

According to the review of the literature about the development of information technologies, by the year 2040, there will be 65 health information technologies (HIT). Earlier records indicate that implementation of these technologies encountered risks and complications in time, cost, and resource allocation which lead to deviation from objectives. In this paper, a strategy is proposed to help organizations to implement future information technologies (IT) by using a balanced scorecard (BSC) and control objectives for information and related technologies (COBIT). Based on COBIT architecture for each technology, there are 4 main domains, 34 related processes, 28 IT goals, and 17 business goals that are introduced within the framework of the BSC. The analytical hierarchy process (AHP) performs the risk assessment for each activity and COBIT framework calculates the maturity of each process. Then the risk and maturity are calculated at the level of processes, domains, and technology. In order to move toward successful implementation of the HIT, results presented as decision-making parameters. These parameters help investors to decide on the allocation of resources based on the risk-taking capability and certainty results and uncertainty results of the plan. The proposed algorithm has advantages over previous methods such as maturity and risk analysis of the organization in information criteria, risk and maturity analysis in management criteria, and calculation of the maturity and risk ratio of business objectives in the current status of the organization.</span>

In a Competitive Market, Understanding Customer Demand and Effective Advertising is one of the most Important Factors in Survival. Extend the Internet and virtual networks have provided a great opportunity for companies to advertise, and thus studying electronic marketing methods and models is of great importance. One of the newest marketing methods is viral marketing that is based on mouth-to-mouth advertising and has a lot of power. Viral marketing relies on the principle that on any social network, a number of users have high power and influence on others, and by identifying them and creating good advertising messages, They can be used to effectively marketing. Therefore, The identification of important users is considered the most important activity in viral marketing. In this regard, various studies have been conducted to identify users using a variety of graph-based and publish-based methods. In this research, the capabilities of both methods have been used and by Using a semi-localized centrality criterion based on graph-based methods and Markov clustering model based on propagation methods, a new hybrid model for user clustering and identification of key users presented. The results show higher correlation between the proposed method and the SIR standard and, therefore, its higher efficiency than other methods used in the research.

With the emergence of Web 2.0, user generated content in the form of online product reviews has proliferated. Although product reviews contain valuable information, they vary greatly in terms of quality and credibility. This study presents an opinion mining framework - Cred-OPMiner (Credibility-Specific-Opinion Miner) - by combining the concepts of credibility and aspect based opinion mining. Cred-OPMiner performs three main tasks. The first task is to group reviewers based on the credibility dimensions. The second critical task is aspect extraction in which aspects of a given product are identified using a novel hybrid and domain independent algorithm. The final task is the sentiment prediction task where the sentiment on each aspect is computed. The key novelty is utilizing source credibility concepts for online reviewer clustering. Source credibility dimensions including trustworthiness and expertise are quantified using reviewers’ data. In addition, a new aspect extraction technique is developed and incorporated in the Cred-OPMiner. Cred-OPMiner was tested using data crawled from epinions.com. It groups reviewers and then performs aspect based opinion mining by differentiating among opinions of various reviewer groups.

A significant amount of treatment cost is paid by health insurance organization. Insurance companies, mostly, use certified people to consider documents, but according to the number of documents and the limitation of time and human resource, consider documents carefully is almost impossible and more importantly, some infringements are not identifiable According to only one document but is identifiable by accumulation of documents and intelligent analysis based on data mining. Detection of beneficial referral (self-referral and kickback) that a doctor refers a patient to a specific pharmacy that has benefits for him, is one of these things.

In this research, data warehouse was prepared by using Tehran health insurance data until 1396 and then after eliminating faulty data, according to network mining methods, actions for detecting anomalistic referrals on the network, data filtering and weighing the edges of the network based on certified people views, were taken. This method was implemented in Knime environment and a short list was presented to health insurance organization’s monitoring department for considering.

In this research, according to the importance of detected interactions during network mining‘s process between doctors and pharmacies, and using visual tools in Knime, 73 doctors were detected that had meaningful relation with 26 pharmacies.

Inspectors of health insurance organization can have a more accurate and more effective examination with spending less time and human resource according to examination patterns based on network mining and visualization.

A key step for success of marketer is to discover influential users who diffuse information and their followers have interest to this information and increase to diffuse information on social media. They can reduce the cost of advertising, increase sales and maximize diffusion of information.  A key problem is how to precisely identify the most influential users on social networks. In this paper, we propose a method to discover influential users based on knowledge management cycle that is called KMIU. The knowledge management cycle consists of several stages including capture, organize, storage, retrieval and mining stages. We try to analyze influential users in two micro bloggings networks as Facebook and twitter by KMIU method. The experimental results showed the proposed method maximize diffusion and has an accuracy 0.55. These maximization and accuracy are more than those of the previous methods.

In the business world and knowledge age, intelligence is one of the undeniable requirements for most organizations so that they can increase their capabilities by increasing deploying knowledge and awareness, and prepare themselves to adapt to changes and environment. Business intelligence is a set of abilities, technologies, tools and solutions that help you better understand business conditions and adapt to it. Data mining is known as a powerful business intelligence tool for discovery, but has not yet fully integrated with it. in the literature, less found a comprehensive and complete approach embracing all aspects of business intelligence. The purpose of this study is to examine the integrated approach that has been reviewed by validated and evaluated models and addresses various aspects of business intelligence, including cost-effective implementation, effective application and system evaluation. It has bees introduced the use of intelligent agents and usiness process data-mining languages In the implementation phase, process of sharing knowledge between the organization's specialists and managers in application phase, and the evaluation of systems in the last phase.

In this paper, the goal of the effort is to determine the future of healthcare and to explain the importance of establishing strategic roadmap in this area. Identification of strategies for future of healthcare includes adaptive and dynamic learning through the industry, governments, and universities. The most important part in generating strategies is to determine the goals with high levels expectations and exclusive path. Healthcare technology goals which are presented in this paper could be inspiring to make a better future. Future developments of medical and health care depend on investing in research, development, and education today. So that health managers achieve goals of the health care system by making appropriate decisions and allocate resources. With appropriate goals and strategies, risk, maturity and reliability level of HIT system could be calculated by using COBIT and best scorecard (BSC) method.

The present study attempts to develop a new model for computing customer lifetime value. The customer lifetime value defined in this paper is the combination of present value and future value. As an innovation the CLV modeling of this paper is based on customer behavior modeling, done by data mining techniques. By extracting the profit vector related to each type of customer behavior, calculation of present study was done, then by utilizing Markov chain model we predict future value and count customer lifetime value. A new churn model was contributed by authors to manage unprofitable CRM costs; utilizing this churn model, the proposed CLV model can cause more profitability for the enterprise. The new CLV model of this paper was validated by historical customer data of a composite manufacturing company.

Knowledge is one of the most important resources of every company that must be effectively managed like any other resources. In addition to inter-organizational reasons, most of companies consider knowledge management and its effective usage as a leverage to be able to compete in their field. Although a lot of research has been conducted in this area, company's manager and decision makers fail in choosing the suitable knowledge management strategy which is the basis and initial step in their field of work. Therefore this research has been done with the goal of selecting the appropriate strategy for project oriented companies. Proposal of this research is to consider the main success factors of knowledge management, besides of its individual properties for each project, and also achievement of their reciprocal relationship through fuzzy DEMATEL method. For this purpose, according to previous studies, 6 main successful factors of knowledge management and 10 important properties of enterprise resource planning implementation project as secondary factors have been identified. Aforementioned project is related to one of the biggest distribution industry companies in Iran. Selected factors serve as assessment criterion for 3 Human-oriented strategy, system-oriented strategy, and dynamic strategy. Finally, by using fuzzy DEMATEL method, importance and efficacy of each strategy according to criteria has been identified.

One of important problems in supply chains management is supplier selection. In a company, there are massive data from various departments so that extracting knowledge from the company’s data is too complicated. Many researchers have solved this problem by some methods like fuzzy set theory, goal programming, multi objective programming, the liner programming, mixed integer programming, analytic hierarchy process (AHP), analytic network process model, TOPSIS, etc. Past research gaps are lack of attention to enterprise historical data and extract knowledge from them, review the past performance of suppliers and use effect of the their past performance to their future work. The aim of this paper is to solve supplier selection problem based on historical data by a novel model. The proposed model has tried to uncover hidden relation in massive unstructured industrial data and has used them to extract knowledge for optimizing decision making and predicting in supply chain management by BI tools. The model is based on FP-Growth algorithm integrated with AHP. Moreover, the proposed model is a multi-criteria decision making model (MCDM) with four criteria: quality, priority, delay on delivery and cost that have chosen from literature review. The criteria have been weighed by AHP and finally the model has been validated by industrial group’s historical data.