کمال میرزایی

This article is an investigation of some problems of modeling, expression of characteristics and analysis of Complex Dynamic Systems (CDS) using Fuzzy Cognitive Maps (FCM). Efficient analysis and control of CDS is impossible without a formal model of the system. The key problem of CDS and control theory is to develop methods to qualitatively analyze the dynamics and behavior of such systems and to construct efficient control algorithms for their efficient operation. Most qualitative description of the parameters of complex dynamic systems inevitably leads to ambiguity, complexity and uncertainty. Reasons for using Fuzzy Cognitive Maps (FCM) in CDS modeling are presented. FCMs provide an opportunity to generate new knowledge based on system applications and address the need to address the uncertainties, ambiguities, and inaccuracies associated with real CDS problems. In this study, the basic principles of FCM are briefly presented and then the NHL algorithm for training the FCM model and adjusting these weights. With fuzzy expressions and converting qualitative data into quantitative, a presentation and an example about the factors affecting the phenomenon of divorce were presented. The example results are promising, further development of CDS modeling and control using soft computing methods.

This research aims to enhance the accuracy of credit assignment for rules in a learning classifier system using Markov Reinforcement Learning for predicting the secondary structure of proteins. Markov Reinforcement Learning has replaced the Bucket Brigade algorithm in the learning classifier system. The Protein Data Bank dataset is utilized to train the system, specifically the protein 4L1W with 5741 samples, where 70% is used for training and 30% for testing purposes. Following the system's training, a set of valuable classifiers (rules) is generated, which will be employed in the testing phase to predict the protein's secondary structure. The experimental results demonstrate an improvement in the accuracy of the Markov Reinforcement Learning classifier system, both in the Learning classifier system and the eXtended classifier system. Through Markov Reinforcement Learning, the credit assignment to each rule is enhanced, resulting in an accuracy improvement of 82.5% for the Learning classifier system and 85% for the eXtended classifier system.

Background &

One of the optimization issues is resource allocation. This means that resources should be allocated in different places so that the cost of allocating the problem is reduced. In addition to its theoretical aspect, the issue of resource allocation can be used in various practical issues such as the timing of project resources, the distribution of foodstuffs in city-level supermarkets, the allocation of medical and emergency centers in a city, etc. Collective intelligence algorithms can be used to solve such problems. In the last decade, various collective intelligence algorithms have been used to solve complex problems such as resource allocation. In this article, the PSO algorithm and its improved variants are used. In the PSO algorithm, it refers to each of the components of a particle. To continue moving, each particle uses its previous experience and that of its neighbors. Every particle in space is moving at a speed. At every moment, this movement leads to the change of the current position of each particle in space. In the PSO algorithm, the communication between particles takes place in two parts of the algorithm. One is the personal best fit of each particle and the other is the best global fit. All the particles that are generated at the beginning of the algorithm as the initial population. Then the fit of each particle is calculated and stored as its personal best fit. In the same way, in updating the position of each particle, this value is compared with the fit of the same particle in the new position, if this fit is better than the previous position of the particle, it is stored as the best personal fit of the same particle. If there is a change in the personal best fit, the fit of this solution is compared with the global best fit. Then, the global best fit is updated.

In this article, the particle swarm optimization algorithm is used to solve various allocation problems. To increase the efficiency, the particle swarm optimization algorithm has been improved with the help of local search. To better strengthen the communication between particles, the concept of a complex network has been used. The status of that particle in the complex network is effective for deciding on the choice of particles. The degree criterion is considered for selecting solutions in the complex network. Two groups of optimization problems have been used to evaluate the results. The first group is different standard and theoretical problems from the QAPLib library for solving the quadratic assignment problem. The second group is also a problem with the location of medical and emergency centers, considering the probability of the center's failure according to real-world conditions. In this issue, it is discussed that according to the criteria of location, cost of construction and reconstruction, and the distance of medical and emergency centers from each other in different population spots, urban medical and emergency centers should be established in which population spot.

In order to compare and evaluate different proposed approaches for resource allocation, two groups of different problems have been used. The first group of standard problems from the QAPLib library is for quadratic assignment. The second group is solving the problem of locating medical and emergency centers in different population spots of the city. The evaluation results of both groups show the average cost and error percentage of the proposed memetic particle swarm optimization algorithm along with the complex network with degree measure compared to the memetic particle swarm optimization algorithm and the basic particle swarm optimization algorithm. For example, to solve the Tia60 problem, the memetic mesh particle swarm optimization algorithm, the memetic algorithm, and the base algorithm have 4.55%, 4.59%, and 8.15% error percentages, respectively.

The results of the implementation of the first to third algorithms for locating medical and emergency centers with 100 repetitions and an initial population size of 50 shows. The memetic degree complex grid particle swarm optimization algorithm has reached the optimal value faster than the standard PSO algorithm and PSO with local search. The third method is more suitable than the first and second method for locating medical and emergency centers. The proposed optimization algorithm of memetic particle swarm with the complex network with degree scale for locating medical and emergency centers has a lower average cost and error percentage than the memetic particle swarm optimization algorithm and basic particle algorithm. One of the problems of optimizing the allocation of resources is the location of medical and emergency centers in urban population spots. This problem has been solved with the help of three different meta-heuristic approaches. To allocate medical and emergency centers, a particle swarm optimization algorithm has been used. Then this algorithm has been converted into a memetic algorithm with the help of the proposed local search. To create a connection between particles, the concept of a complex network has been used. In the third proposed method, in addition to the particle swarm optimization algorithm, along with local search, a complex network with a degree scale is used. To evaluate the results, two groups of problems have been used. The first group has been used to solve various quadratic assignment problems from the QAPLib library. In the second group, the practical problem of locating medical and emergency centers of different population spots in the city is considered. To solve both groups, the proposed memetic complex network particle swarm optimization algorithm has a lower average cost and error percentage than the memetic particle swarm optimization algorithm and the basic particle swarm optimization algorithm. To improve the performance of the proposed algorithms, other metrics can be used in the complex network or he used other local search algorithms such as refrigeration simulation, and forbidden search.

In this paper, Hamiltonian cycle was used in a standard and theoretical problem called TSP as well as a practical problem called finding the shortest Hamiltonian cycle to cover all provinces of Iran. These kinds of problems can be solved using swarm intelligence algorithms derived from natural, environmental and social factors. Accordingly, the PSO Algorithm, which is one of the algorithms of swarm intelligence, was used . Search was also used to improve the results of each particle. Besides, a complex network was used in order to enhance the better exchange of information between particles and to select the next most appropriate position for each particle. In this network of nodes in which a better solution is maintained, the degree of that node is always greater. In the complex network, two criteria, degree and neighborhood degree, are used to find the best solution. The TSPLib standard problems were used to compare the results.  The findings showed that the particle swarm optimization technique with a complex network local search was more cost-effective than the optimization of the particle swarm with local search and standard particle swarm in a way that the percentage of error to the best solution in TSPLib was reduced in the particle swarm optimization algorithms with complex network local search and the particle swarm optimization in the local search compared to the standard particle swarm method, respectively. To solve the ST70 in network and basic algorithms, the average cost of solving the problem is 705 and 797, respectively.

An ensemble clustering has been considered as one of the research approaches in data mining, pattern recognition, machine learning and artificial intelligence over the last decade. In clustering, the combination first produces several bases clustering, and then, for their aggregation, a function is used to create a final cluster that is as similar as possible to all the cluster bundles. The input of this function is all base clusters and its output is a clustering called clustering agreement. This function is called an agreement function. Ensemble clustering has been proposed to increase efficiency, strong, reliability and clustering stability. Because of the lack of cluster monitoring, and the inadequacy of general-purpose base clustering algorithms on the other, a new approach called an ensemble clustering has been proposed in which it has been attempted to find an agreed cluster with the highest Consensus and agreement. In fact, ensemble clustering techniques with this slogan, the combination of several poorer models, is better than a strong model. However, this claim is correct if certain conditions (such as the diversity between the members in the consensus and their quality) are met. This article presents an ensemble clustering method. This paper uses the weak clustering method of fuzzy cmeans as a base cluster. Also, by adopting some measures, the diversity of consensus has increased. The proposed hybrid clustering method has the benefits of the clustering algorithm of fuzzy cmeans that has its speed, as well as the major weaknesses of the inability to detect non-spherical and non-uniform clusters. In the experimental results, we have tested the proposed ensemble clustering algorithm with different, up-to-date and robust clustering algorithms on the different data sets. Experimental results indicate the superiority of the proposed ensemble clustering method compared to other clustering algorithms to up-to-date and strong.

Given the growing popularity and use of digital devices in human daily life; an increase in image sharing on social networks such as Facebook , Flicker, Instagram as well as loading various clips on such networks , the utilization of digital images has sharply increased, particularly in the last decade, among them, a high percentage belongs to human face images(approximately above 60%) what matters here is face image retrieval using high accuracy feature extraction methods. The more precise method, the more time it requires to be processed. in this paper we plan to execute numerous high accuracy feature extraction methods in face image retrieval like LBP and its variants, HOG, SIFT, SURF and in some case the combination of above mentioned methods by suitable coding on GPU in order to examine the impact of processing speed and its accuracy compared to the CPU and propose the best method for it

Vector Quantization (VQ) is the powerful technique in image compression. Generating a good codebook is an important part of VQ. There are various algorithms in order to generate an optimal codebook. Recently, Swarm Intelligence (SI) algorithms were adapted to obtain the near-global optimal codebook of VQ. In this paper, we proposed a new method based on a modified firefly algorithm (MFA) to construct the codebook of VQ. The proposed method merged genetic crossover operator with FA to develop the VQ. This method is called MFA model. Experimental results indicate that the reconstructed images generated by the proposed model is get higher quality than FA and it’s about one percent, but it is no significant superiority to the PSO algorithm. Furthermore, MFA is slower than FA.

The potential battle scene against the Islamic Republic of Iran (I.R.I) is network-centric operation with Effect Based approach. One of the I.R.I capabilities in this complex battle scene is the capabilities of the cyber-electronics. One of the major challenges of planning cyber-electronic operations for confrontation in such a battle scene in the military decision-making process, is comparing and selecting courses of actions (COAs). The multiplicity, interdependencies, direct and indirect effects of the network centric operation criteria have made comparison and selection of COAs, a complex problem. In this paper, the decision model for choosing an operational solution with consideration of the complexity of the problem is presented based on hybrid methods DANP and TOPSIS. The proposed model consists of three steps.  In the first step, valid data produced by selected experts is used as data input. In the second step, data processing involves calculating an influential criteria network (which models the influence of criteria) and finding the influential weights of DANP (DEMATEL-based ANP) considering the relationships between them. The third stage of evaluation and selection is based on the level of similarity to the ideal solution (TOPSIS). In order to evaluate the proposed model, the real scenario of network centric operation of the US Operation Iraqi Freedom (OIF) from March 21 to 22, 2003 was evaluated. In this operation, the interactions between the 5th Corps command posts (180 posts) were identified as the most important criteria. The quality of force networking based on FBCB2 / BFT systems, and mission effectiveness were determined respectively as the most effective and the most influenced criteria on the battlefield. In the second step, the relative weights of the criteria were determined assigning the highest weights to agility and effectiveness and the lowest weight to the quality of information. Among the cyber-electronics solutions offered by the experts, the L-satellite Jamming and the injection of electronic malicious code, were the effective ways to confront the 5th US  Army Corps mission to capture Talil air base. The selected expert commanders declared the results of this case study as valid and satisfactory.

US Network Centric Warfare is one of the threats lying ahead of the Islamic Republic of Iran. To deal with such a threat, recognizing the Network Centric Warfare (NCW) and its main criteria are fundamental issues that must be examined from different perspectives. The multiplicity of interdependencies, feedbacks and direct and indirect interactions between components and criteria of network centric operations, has turned the comparison and evaluation to a complicated problem. To tackle this complex problem, in this paper we used DEMATEL's method. The level of influence, impact, degree of importance, and causal     relationship of criteria to each other are evaluated. Based on data analysis, individual sensemaking and quality of information, are the most influential criteria. The degree of shared sensemaking is the most     effective criterion and agility, individual sensemaking and synchronization follow next in rank. In terms of the importance of criteria that are somehow centroid in terms of impact, sensemaking has the highest rank and then the quality of information and interactions between the forces and their allies have the next rank. From the perspective of the cause and effect relationship, the quality of information criteria has the highest rank as a " cause"  and effectiveness and agility as " effects" .The achievements of this strategic assessment can be focused on prioritizing active and passive regions in the network centric battle scene. Confirming the cyber-electronic experts’ opinion, another important achievement is the effectiveness of the role and status of cyber-electronic capabilities in the network centric operation battle scene, which needs to be assessed accurately. Simulation and data analysis have been done in MATLAB software.

The prediction of future battles is a subject that has been debated between theorists for decades . According to them, this battlefield is network-based, complex, non-linear, dynamic, and in an uncertain environment. Since, based on the theory of complexity and chaos, minor problems can have destructive effects in this complex environment, its evaluation of damage is a research necessity. The research methodology is based on document analysis and library study to extract the damage and assessment criteria. Then, based on the hierarchical analysis process, we present a model for assessing and ranking the network-based operation damages. Vulnerability assessment criteria are time, cost, effectiveness, and feasibility, for evaluating vulnerabilities for effective use.Based on evaluation criteria, electronic warfare, technology and command vulnerabilities are the top priority for use. Vulnerabilities, such as electronic warfare, technology and human resources are with the highest rating in terms of time. Based on cost, network and communications, technology and electronic warfare vulnerabilities have the highest priority and command vulnerabilities such as erosion of the war and over loading information have the lowest cost rating. The most effective vulnerability of science progress includes theories of complexity, chaos and system theory .The least-effective vulnerability is due to technology. Based on feasibility criteria, command damage, and information risk have the highest rank, which can be achieved with psychological and cognitive operations. The threats of information deception and the development of science were the lowest rank.

Gastric cancer is the second leading cause of cancer death in the world. Due to the prevalence of the disease and the high mortality rate of gastric cancer in Iran, the factors affecting the development of this disease should be taken into account. In this research, two data mining techniques such as Apriori and ID3 algorithm were used in order to investigate the effective factors in gastric cancer.
Data sets in this study were collected among 490 patients including 220 patients with gastric cancer and 270 healthy samples referred to Imam Reza hospital in Tabriz. The best rules related to this data set were extracted through Apriori algorithm and implementing it in MATLAB. ID3 algorithm was also used to investigate these factors.
The results showed that having a history of gastro esophageal reflux has the greatest impact on the incidence of this disease. Some rules extracted through Apriori algorithm can be a model to predict patient status and the incidence of the disease and investigate factors affecting the disease. The prediction accuracy achieved through ID3 algorithm is 85.56 which was a very good result in the prediction of gastric cancer.
Using data mining, especially in medical data, is very useful due to the large volume of data and unknown relationships between systemic, personal, and Behavioral Features of patients. The results of this study could help physicians to identify the contributing factors in incidence of the disease and predict the incidence of the disease.

Decision tree is the data mining tools to collect, accurate prediction and sift information from massive amounts of data that are used widely in the field of computational biology and bioinformatics. In bioinformatics can be predict on diseases, including breast cancer. The use of genomic data including single nucleotide polymorphisms is a very important factor in predicting the risk of diseases. The number of seven important SNP among hundreds of thousands genetic markers were identified as factors associated with breast cancer. The objective of this study is to evaluate the training data on decision tree predictor error of the risk of breast cancer by using single nucleotide polymorphism genotype.
The risk of breast cancer were calculated associated with the use of SNP formula:xj = fo * In human, The decision tree can be used To predict the probability of disease using single nucleotide polymorphisms .Seven SNP with different odds ratio associated with breast cancer considered and coding and design of decision tree model, C4.5, by Csharp2013 programming language were done. In the decision tree created with the coding, the four important associated SNP was considered. The decision tree error in two case of coding and using WEKA were assessment and percentage of decision tree accuracy in prediction of breast cancer were calculated. The number of trained samples was obtained with systematic sampling. With coding, two scenarios as well as software WEKA, three scenarios with different sets of data and the number of different learning and testing, were evaluated.
In both scenarios of coding, by increasing the training percentage from 66/66 to 86/42, the error reduced from 55/56 to 9/09. Also by running of WEKA on three scenarios with different sets of data, the number of different education, and different tests by increasing records number from 81 to 2187, the error rate decreased from 48/15 to 13/46. Also in the majority of scenarios, prevalence of the disease, had no effect on errors in the WEKA and code.
The results suggest that with increased training, and thus the accuracy of prediction error decision tree to reduce the risk of breast cancer increases with the use of decision trees. In Biological data, decision trees error is high even with a 66/66% training. On the other hand by increasing the number of SNP from 4 to 7 decision tree, decision tree error dramatically decreased at 70/1% training. In general we can say that with increased training and increasing the number of SNP in the decision tree, the prediction accuracy increased and errors reduced. In the CODING and WEKA, percentage of disease prevalence had no significant effect on errors,” Because of selecting set of training and testing by systemic method “.

Gastric cancer is the second cause of death from cancer after lung cancer in the world It`s incidence is varied in different regions of the words. Due to the prevalence rate of the disease and high mortality rates for gastric cancer in the country, it is necessary to examine the influential factors in the incidence of this disease by more accurately and scientific methods. The purpose of this study is to examine this factor with data mining techniques. The required data for this study was collected from patients referring to Imam Reza Hospital, Tabriz. After applying data pre-processing, totally 490 records were collected in an Excel file samples including 220 cancer cases and 270 normal specimens. The best rules based on the datasets were extracted by using Apriori algorithm implemented in MATLAB software and final data set. In this study, gastric cancer datasets and features affecting the incidence of this disease have been used for the first time. The results showed that risk for gastric cancer in people with cardiovascular disease are less. In addition, gastric reflux is associated with not using salt and milk, and high salt intake. Gastric reflux has also the most influence on creating this disease. Some rules were obtained by using Apriori algorithm that can be used as a model to predict the status of patients and the incidence of this disease. Nowadays Due to massive amounts of medical data, knowledge can be extracted from datasets by using data mining approach. In this study, some rules were extracted by using Apriori algorithm that can provide physicians with great help to examine the causes of this disease.

Study of yield of some agronomic and physiological traits of barley cultivars and lines under salinity stress conditions