A few artificial neural networks have been proposed so far for multivariate time series prediction and they used simple general-purpose neural networks. Therefore, they cannot achieve high prediction accuracy. In this paper, we propose an artificial neural network called DBMTSP (Dependency Based Multivariate Time Series Prediction) to predict the next element of a time series in the multivariate case. Compared to the existing methods, DBMTSP considers both intra-time-series dependencies and inter-time-series dependencies efficiently to achieve more accurate predictions. We propose a hierarchical encoder in DBMTSP to discover inter-time-series dependencies. The proposed hierarchical encoder is able to encode secondary time series into a single parameter that represents dependencies that exist between the main time series and the secondary time series. The hierarchical encoder has a scalable design such that it can accept a large number of secondary time-series. We have trained DBMTSP using 32760 data matrices. We evaluated DBMTSP using 8190 test data matrices. Our evaluations show that DBMTSP surpasses the existing methods in term of prediction accuracy.
Web development goal was information resources sharing between individuals and organizations. While resources is increasing, our processing capacity about a particular subject remains unchanged. Decision making in any field with available information resources on the web has a significant impact on human life. In this paper, an AHP method is proposed which pays attention to the concept of trust dimensions for available information resources credibility evaluation on the web. Our proposed method based on the existing ontology with a semantic search engine extracts various attributes to evaluate the credibility of a resource and assigns a numerical value using experimental evaluation. To confirm the correctness of the proposed method, an experiment has been organized to select a book based on the author's academic degree, author's reputation, publication year, publisher's reputation, organization and structure, price and etc. The feature ranking of mentioned book is performed, and finally, achieved results are expressed.
Strategy map is a human based activity that combines the knowledge and organizational managers’ opinions. This point of view is a kind of group decision making that is determined by our existing relation between the accepted objectives of experts and organizational decision makers (DM). In this study, to design a strategy map for Fanap Company, the combined approach of Analytical Network Process (ANP) and Decision Making Trial and Evaluation (DEMATEL) were applied to determine the systematic interactions between criteria and the relationship between criteria, respectively. In this paper, the results about the importance of relations between strategic objectives level by combined approach, were compared to the importance of relations between AHP strategic objectives level, calculated by Fanap Company’s experts and managers. The final results showed that the importance of strategic relations between BSC levels to achieve the strategic objectives in this company in combined approach was more than AHP method.
Recently, Combined Heat and Power (CHP) systems have been utilized increasingly in power systems. With the addition penetration of CHP-based co-generation of electricity and heat, the determination of economic dispatch of power and heat becomes a more complex and challenging issue. The optimal operation of CHP-based systems is inherently a nonlinear and non-convex optimization problem with a lot of local optimal solutions. In this paper, the Improved Shuffled Frog Leaping Algorithm (ISFLA) is used for solution of the problem. ISFLA is an improved version of shuffled frog leaping algorithm in which new solutions are produced in respect to global best solution. The ISFLA is well able to attain the optimal solutions even in the case of non-convex optimization problems. To evaluate the efficiency of the proposed method, it has been implemented on the standard test system. The obtained results have been compared with other heuristic methods. The numerical results show that the ISFLA is faster and more precise than other methods.
Domain adaptation is a powerful technique given a wide amount of labeled data from similar attributes in different domains. In real-world applications, there is a huge number of data but almost more of them are unlabeled. It is effective in image classification where it is expensive and time-consuming to obtain adequate label data. We propose a novel method named DALRRL, which consists of deep architecture with domain-general and domain-specific representations across domains for deep unsupervised domain adaptation. Also, we apply low-rank representation learning to reduce source and target domains discrepancy. The low-rank constraint can uncover more related information between domains and it can transfer more relevant knowledge from the source domain to the target domain. The DALRRL guarantees to minimize marginal and conditional distributions difference between the source and target domains. The experimental results conducted on two benchmark domain adaptation datasets demonstrate the effectiveness of our method in image classification tasks.
Recommender systems are the systems that try to make recommendations to each user based on performance, personal tastes, user behaviors, and the context that match their personal preferences and help them in the decision-making process. One of the most important subjects regarding these systems is to increase the system accuracy which means how much the recommendations are close to the user interests. In this paper, to achieve the mentioned aim we use a combination of K-means and differential evolution algorithms. The K-means algorithm determines the best recommendations for the current user based on the behavior of the other users. The differential evolution algorithm is used to optimize the user clustering in the recommender system. Given that the proposed model has been tested in a movie domain, the films suggested to the current user, have the highest rates from the users who are similar to the current user. The results gained from the simulation show the superior performance of the proposed model in comparison to the related works with an average increased accuracy of 0.01.
A five-level Power Factor Correction incremental rectifier (PFC) is proposed in this paper. In this topology, the output voltage and current of the rectifier are controlled using the multilevel modulation and smart controller technologies. A multi-carrier pulse width modulation is used to create the switching pulse. In this topology, the number of semiconductor switches is reduced to 3. The smart controller is implemented using a Multilayer Perceptron (MLP) neural network and it is trained using the backpropagation algorithm. This controller is used instead of the well-known PID controller to control the input voltage and current. It should be noted that in this work, the goal is to design an intelligent controller using a neural network instead of a PID controller. The results obtained using this controller as compared to the PID controller show a decrease in the peak voltage, an increase in the rise time, and a ripple reduction in the output voltage. This study is conducted using the Simulink environment in MATLAB and the results suggest that a smart controller can be an alternative to the PID controller.
Flexible AC Transmission System (FACTs) devices are used in power transmission networks to increase maximum power transmission and stability. On the one hand, they help to damp low-frequency oscillations for both local and internal areas. But on the other, the design of these devices with uncoordinated Power System Stability (PSS) may degrade the performance of the power system. In addition, the power systems are vast, complex, and nonlinear. Linear control strategies do not have satisfactory performance for these systems, especially when some disturbances occur. In this study, a nonlinear Coordinated Control Strategy based on Recurrent Neural Network (CCSRNN) is designed to control PSS and Static Synchronous Compensators (STATCOM) coordinates for a standard multi-machine power system and also using the Multi-Bound Power System Stability (MB-PSS) and Multi-Bound Recurrent Neural Network MB-RNN PSS and compare the result of each one for having better stability of the system. The results simulation proves that the controller leads to transient stability and low-frequency oscillation.
Today, the need to deal with ambiguous information in semantic web languages is increasing. Ontology is an important part of the W3C standards for the semantic web, used to define a conceptual standard vocabulary for the exchange of data between systems, the provision of reusable databases, and the facilitation of collaboration across multiple systems. However, classical ontology is not enough to handle the vague information commonly found in many practical areas. An acceptable solution is to incorporate the ability of fuzzy logic to extend classical ontology. SWRL is used to build fuzzy ontology classes and property maps, as well as computations; it is added to ontology and creates new knowledge. In this paper, we introduce an approach based on the production of the fuzzy ontology using SWRL rules, which are the rule-writing language on the Semantic Web. Accordingly, a method for combining SWRL and OWL to produce a fuzzy ontology is proposed. Fuzzy logic is a way to model complex systems that are impossible or very difficult to model using classical modeling methods, much more easily and flexibly.
Induction motors are one of the most widely used machines that are used in industrial motion control and home systems. On the other hand, the optimal determination parameters have a direct and significant effect on their efficiency, longevity, and performance. Determining the initial parameters of these machines by the classic method is time-effective and costly. Therefore in recent years, the use of computer processing such as simulation, heuristical algorithm, ANN, etc. has become common as an alternative method in this field. Cuckoo Optimization Algorithm (COA) is a relatively new algorithm and it is less used to solve this nonlinear problem. In this article, the parameter of the induction machine will be estimated using COA. COA. In the following, the results obtained from the algorithm will be compared with the results obtained from several examples of conventional algorithms. Objective functions are defined as minimizing the true values of the relative error between the measured and estimated torques of the machine in different slips. Objective functions have been used to optimize the parameters of the induction motor with an approximate equivalent circuit and the induction motor with the exact circuit. The final results show the accuracy of operation and very good convergence speed of the algorithm in solving the problem.
Text coherence evaluation becomes a vital and lovely task in Natural Language Processing subfields, such as text summarization, question answering, text generation and machine translation. Existing methods like entity-based and graph-based models are engaging with nouns and noun phrases change role in sequential sentences within short part of a text. They even have limitations in global coherence evaluation, especially in long and narrative documents. This paper presents a new and simplified method for evaluating local and global text coherence. The proposed method focuses on entity grid method and employs two graph-based and entropy-based approaches to overcome its challenges and shortcomings. Applying statistical approaches, the presented method studies how to incorporate other entity properties into short and long stories to assess both local and global coherence, simultaneously. Results indicate that the proposed method is superior to other algorithms in terms of performance, accuracy in long documents with a high number of sentences.
Data clustering is an ideal way of working with a huge amount of data and looking for a structure in the dataset. In other words, clustering is the classification of the same data; the similarity among the data in a cluster is maximum and the similarity among the data in the different clusters is minimal. The innovation of this paper is a clustering method based on the Crow Search Algorithm (CSA) and Opposition-based Learning (OBL). The CSA is one of the meat-heuristic algorithms that is difficult at the exploration and exploitation stage, and thus, the clustering problem is susceptible to initialization for centrality of the clusters. In the proposed model, the crows change their position based on the OBL method. The position of the crows is updated using OBL to find the best position for the cluster. To evaluate the performance of the proposed model, the experiments were performed on 8 datasets from the UCI repository and compared with seven different clustering algorithms. The results show that the proposed model is more accurate, more efficient, and more robust than other clustering algorithms. Also, the convergence of the proposed model is better than other algorithms.
Morphology has a special place in any language, including written and spoken applications. Markov method is used to labeingl and determine the role of words.emergence in software sciences has eliminated 0 and 1 computations, putting them within an infinite space of between 0,1. This characteristic of fuzzy logic has resolved ambiguity in numerous previous problems. The sentence roles in Persian language were specified based on the fuzzy logic’s capability to resolve ambiguity. In two defuzzification methods Mean Of Max, Central Average, the role of words in the sentence is identified and the success rate of each method is obtained. Finally, Mean Of Max with a success rate of 64% proved to be a defuzzifier delivering the best output among two different defuzzification methods. ** * * * * * * * * * * * * * * * * * * ****** * * * * * * * * *
Recently by developing the technology, the number of network-based servicesis increasing, and sensitive information of users is shared through the Internet.Accordingly, large-scale malicious attacks on computer networks could causesevere disruption to network services so cybersecurity turns to a major concern fornetworks. An intrusion detection system (IDS) could be considered as anappropriate solution to address the cybersecurity. Despite the applying differentmachine learning methods by researchers, low accuracy and high False AlarmRate are still critical issues for IDS. In this paper, we propose a new approach forimproving the accuracy and performance of intrusion detection. The proposedapproach utilizes a clustering-based method for sampling the records, as well asan ensembling strategy for final decision on the class of each sample. For reducingthe process time, K-means clustering is done on the samples and a fraction of eachcluster is chosen. On the other hand, incorporating three classifiers includingDecision Tree (DT), K-Nearest-Neighbor (KNN) and Deep Learning in theensembling process results to an improved level of precision and confidence. Themodel is tested by different kinds of feature selection methods. The introducedframework was evaluated on NSL-KDD dataset. The experimental results yieldedan improvement in accuracy in comparison with other models
Data Grid provides sharing services for very large data around the world. Data replication is one of the most effective approaches to reduce access latency and response time. In addition to the benefits, replication has costs such as storage and bandwidth consumption, especially when storage space is low and limited. Therefore, the data replacement should be done wisely. In this paper, we proposed a replacement method called FRA. The algorithm defines a weight for each replica that represents its value. This algorithm uses this weight to prevent the removal of valuable replicas. The results demonstrated that FRA algorithm has better performance than other replication methods in terms of the number of replications, the percentage of storage used, and the job execution time.
Recently, fog computing has been introducedto solve the challenges of cloud computing regarding Internet objects. One of the challenges in the field of fog computing is the scheduling of tasks requested by Internet objects. In this study, a review of articles related to task scheduling in fog computing has been done. At first, the research questions and goals will be introduced, and then we will explain the process of finding and selecting the articles. A comprehensive analysis of the articles will be done. We have identified and listed 10 optimization metrics. Also, according to our study, in 79% of the studied articles, the mathematical model was used to express the problem. In 42% of the articles meta-heuristic algorithms proposed and 84% evaluated their algorithm by simulation. Finally, this paper presents the challenges and open issues of task scheduling in fog computing to the researchers.
Underwater Wireless Sensor Networks (UWSNs) are considered as wireless sensor networks whose main task is to sense underwater events and send information to the sink. This information becomes valuable when the exact location of the occurrence is known. Generally, underwater sensor nodes are not equipped with devices such as the Global Position System (GPS) with the purpose of reducing network costs. Therefore, finding the location of the nodes should be done using another exact method. In this paper, we intend to find the location of the underwater sensor nodes by introducing a new method based on the Cuckoo Optimization Algorithm (COA). We will compare the proposed method with the related methods in terms of the localization error rate and the number of nodes discovered. The results of the comparisons show that the proposed method can greatly reduce the error rate of the localization of the sensor nodes.
Convolutional neural networks show outstanding performance in many image processing problems, such as image recognition, object detection and image segmentation. Semantic segmentation is a very challenging task that requires recognizing, understanding what's in the image in pixel level. The goal of this research is to develop on the known mathematical properties of the soft-max function and demonstrate how they can be exploited to conclude the convergence of learning algorithm in a simple application of image recognition in supervised learning. So, we utilize results from convex analysis theory which associated with hierarchical architecture to derive additional properties of the soft-max function not yet covered in the existing literature for Multi-Class Classification problems. The proposed MC-DEEP model represents an average accuracy of 90.25% in different layers setting with 95% confidence interval in best initial settings in deep convolutional layers which applied on MNIST dataset. The results show that the regularized networks not only could provide better segmentation results with regularization effect than the original ones but also have certain robustness to noise.
In this paper, a novel filter-based approach is proposed using the PageRank algorithm to select the optimal subset of features as well as to compute their weights for web page classification. To evaluate the proposed approach multiple experiments are performed using accuracy score as the main criterion on four different datasets, namely WebKB, Reuters-R8, Reuters-R52, and 20NewsGroups. By analyzing the obtained results, it is observed that the accuracy score of the classifier on WebKB, Reuters-R8, and Reuters-R52 datasets significantly improved from 91% up to 96% compared to the best result achieved by other feature selection methods like IG and Chi-2. Whereas, the accuracy score of the classifier on 20NewsGroups dataset didn't see any noticeable improvement and remained close to the most compared methods. Evaluating the performance of the proposed approach shows the superiority of it in obtaining higher accuracy scores when compared with the feature sets selected by other methods.
A main challenge arisen regarding the application of web services is the discovery of a suitable web service for different requirements. After the emergence of Semantic Web and then the emergence of Semantic Web Service, discovering the web services became specifi-cally important. The core ontology is a semantic web and may be used to facilitate the pro-cess of the web services discovery. The aim of web service discovery is to seek and find the web services that can meet the needs of the user. The process of web service may include the combination of several web services if a service isn’t solely capable of meeting the user’s needs. Discovering the web service has been considered extensively by the research associations. The techniques provided in this field could be summarized in syntax and semantic categories. In regard to the suggested methods for semantic service, only the inputs and outputs have been considered. This resulted in the service discovery distraction and consequently the de-sired service did not correspond with the user’s request. The algorithm presented in this re-search considers not only the inputs and outputs of the services but also the precondition and its impact and this is a big advantage in relation to the other service discovery algorithm. The efficiency and rate of precision of given algorithm have been evaluated by the third version of the dataset OWLS – TC and compared with three algorithms of the service discovery with a high precision.
