فهرست مطالب محبوبه شمسی

In recent years, the massive growth of generated content by users in social networks and online marketing sites, allows people to share their feelings and opinions on a variety of opinions about different products and services. Sentiment analysis is an important factor for better decision-making that is done using natural language processing (NLP), computational methods, and text analysis to extract the polarity of unstructured documents. The complexity of human languages and sentiment analysis have created a challenging research context in computer science and computational linguistics. Many researchers used supervised machine learning algorithms such as Naïve Bayes (NB), Stochastic Gradient Descent (SGD), Support Vector Machine (SVM), Logistic Regression (LR) Random Forest (RF), and deep learning algorithms such as Convolution Neural Network (CNN) and Long Short-Term Memory (LSTM). Some researchers have used Dictionary-based methods. Despite the existence of effective techniques in text mining, there are still unresolved challenges. Note that user comments are unstructured texts; Therefore, in order to structure the textual inputs, parsing is usually done along with adding some features, linguistic interpretations and removing additional items, and inserting the next terms in the database, then extracting the patterns in the structured data and finally the outputs will evaluate and interpret. The imbalance of data with the difference in the number of samples in each class of a dataset is an important challenge in the learning phase. This phenomenon breaks the performance of the classifications because the machine does not learn the features of the unpopulated classes well. In this paper, words are weighted based on the prescribed dictionary to influence the most important words on the result of the opinion mining by giving higher weight. On the other hand, the combination of the adjacent words using n-gram methods will improve the outcome. The dictionaries are highly related to the domain of the application. Some words in an application are important but in mobile comments are not impressive. Another challenge is the unbalanced train data, in which the number of positive sentences is not equal to the number of negative sentences. In this paper, two ideas are applied to build an efficient opinion mining algorithm. First, we build a precise dictionary for mobile Persian comments, and the second idea is to balance the positive and negative comments in train data. In summary, the main achievements of the current research can be mentioned: creating a weighted comprehensive dictionary in the field of mobile phone opinions to increase the accuracy of opinion analysis, balancing positive and negative opinions to improve the accuracy of opinion analysis, and eliminating the negative effect of overfitting and providing a precise approach to Determining the polarity of users' opinions about mobile phones using machine learning and recurrent deep learning algorithms. This new method is presented on mobile phone products from the Digikala site and Senti-Pers data. The result is performed with Naive Bayesian, Support Vector Machine, Stochastic Gradient Descent, Logistic Regression, Random Forest, and deep learning methods such as Convolutional Neural Network and Long Short-Term Memory based on parameters such as Accuracy, Precision, Retrieval, and F-Measure. The proposed method increases accuracy on Digikala, with NB between 10% and 34% and SVM between 5% and 24%, SGD between 7% and 38%, LR between 5% to 38%, and RF between 4% Up to 22% and CNN by 4%. The results show an accuracy increment on Senti-Pers, with NB between 12% and 46% and SVM between 5% and 46%, SGD between 5% and 35%, LR between 6% to 46%, and RF between 4% Up to 46%.

In a wireless sensor network (WSN), due to various factors such as limited power, sensor transferability, hardware failure and network problems such as packet collisions, unreliable connection and unexpected damage, the amount sensed to the header or base station is not Arrives. Therefore, data loss is very common in wireless sensor networks. Loss of measured data greatly reduces WBAN accuracy. Because WBAN deals with the vital signs of the human body, network reliability is very important. To solve this problem, missing data must be estimated. In order to predict the missing values, a model for estimating lost data based on LSTM (short-term memory) neural network is presented in this paper. This model combines five vital signs as input to predict the amount lost. The results show that sgdm-LSTM is a good way to estimate the amount lost. In addition, experimental results show that the mean square root error of the estimated value is lower than other methods. This value is 4.1495 with the best network parameters.

In recent years, convolutional networks have achieved impressive results in the field of machine vision. Their high accuracy has led to their use in security and military fields, such as crime prediction and prevention. However, due to the huge calculations required by these networks, they cannot be used in embedded systems with limited computing resources. One solution to this problem is network pruning. In this method, a portion of the network is pruned while maintaining the initial structure to reduce calculations and enable use in embedded systems. Traditional pruning methods typically require manual determination of the criteria and amount of pruning, and they tend to perform well on specific networks. Automatic methods, on the other hand, do not have these disadvantages. However, they must address issues such as relaxing the search space, eliminating overfit during the training process, and addressing differences resulting from relaxing the search space.This article introduces a method based on gradient decent for network pruning. The method automatically searches for the optimal pruned network and can be easily applied to networks with skip connections. To compensate for lost accuracy, dynamic skip connections are used without increasing computation. Tests on valid datasets demonstrate that this method achieves a better balance between accuracy and network pruning than previous methods.

Early diagnosis of breast cancer plays a crucial role in treating the patient. Nowadays, data mining algorithms can provide intelligent methods in the health and treatment system that accurately detect breast cancer. The purpose of this study is breast cancer detection using ensemble classifier based on WBC and WDBC prepared databasesa. Our proposed model in the WBC database (reducing features by cfs+ optimizing samples using Resample+ ensemble classifier using data mining algorithms (kstar + random forest + Naïve Bayes and Bayes network)) has the best detection accuracy ( 100%), implementation time (0 seconds) and without any errors and on the WDBC database (reducing features by cfs+ optimizing samples using Resample+ ensemble classifier using data mining algorithms (IBK algorithm+ Naïve Bayes, Bayes network and kstar)) has an accuracy of 99/29, the implementation time is 0 seconds, and the mean absolute error is 0/007. The results of this study show that according to the ensemble classifier methods using data mining algorithms on the prepared database, new systems can be designed to help physicians that facilitate treatment processes.

Classification of cancer subtypes is very important task for the diagnosis and prognosis of cancer. In recent years, deep learning methods have gained considerable popularity for this reason; however, it is difficult to determine the structure of the neural network because the function of the deep network depends largely on its structure. In addition, the high number of genes in the gene expression database and the imbalanced data between different classes have a direct effect on the complexity and performance of cancer subgroup classification models. To address the problem of unbalanced data, a convolution neural network (CNN) model using a cost-sensitive strategy is proposed to increase the model's accuracy in identifying minority classes. On the other hand, the fisher ratio technique is used to reduce genes in the preprocessing stage. In techniques the cost-sensitive method, a cost matrix is created based on the distribution of classes, and then this matrix is used in the CNN network cost function step to calculate the amount of error. Two sets of cancer datasets are used to evaluate the proposed method. The results show that selecting the appropriate genes for classification along with the use of cost-sensitive learning can increase the performance of the proposed method compared to the CNN model without selecting the feature and cost-sensitive learning about 11%, 10% and 18% in terms of three criteria of accuracy, recall and precision, respectively.

In recent years, the number of sick and elderly people that are living alone in their homes and need care has grown. This is why the smart home is needed for the awareness of their condition. Identification of the patient's activity using environment sensors is the first step in implementing a smart home. In such a home, the patient's relatives can leave the patient alone with less concern. In this research, various methods of recognizing user’s activity in the smart home are presented, and then a new method is introduced to diagnose the risk amount of Alzheimer's patients in which fuzzy-logic is used in cases such as start-up time of the activity. This system uses fuzzy logic in three phases. Since this method is considered for specific patients, the wearing sensors are not used because they have difficulties for the elderly and even an Alzheimer's patient may forget to wear them. However, the implementation of this layer also achieved good results, i.e., 84% accuracy.

Early diagnosis of breast cancer plays a crucial role in treating the patient. Nowadays, data mining algorithms can provide intelligent methods in the health and treatment system that accurately detect breast cancer. The purpose of this study is breast cancer detection using ensemble classifier based on WBC and WDBC prepared databasesa. Our proposed model in the WBC database (reducing features by cfs+ optimizing samples using Resample+ ensemble classifier using data mining algorithms (kstar + random forest + Naïve Bayes and Bayes network)) has the best detection accuracy ( 100%), implementation time (0 seconds) and without any errors and on the WDBC database (reducing features by cfs+ optimizing samples using Resample+ ensemble classifier using data mining algorithms (IBK algorithm+ Naïve Bayes, Bayes network and kstar)) has an accuracy of 99/29, the implementation time is 0 seconds, and the mean absolute error is 0/007. The results of this study show that according to the ensemble classifier methods using data mining algorithms on the prepared database, new systems can be designed to help physicians that facilitate treatment processes.

Prediction of flow filed in compound channel is an important task for hydraulic researches because of three-dimensional nature of flow. The complexity of the problem increases, significantly, when floodplains geometry changes from prismatic to non-prismatic form. In non-prismatic compound channels with converging floodplains, the main feature consists of the mass and consequent momentum exchange between the floodplains and the main channel. In the present work based on the depth-averaged Navier-Stokes equation an analytical method is proposed and used to predict the depth-averaged velocity and boundary shear stress distribution in non-prismatic compound channels with converging floodplains.  In order to consider the effect of the secondary flows in the depth-averaged Navier-Stokes equation, temporal mean velocity components are assumed to be a portion of the depth-averaged velocity. Also, since the flow condition is not uniform; the flume bed slope has been replaced by the energy line slope. The results of the proposed analytical method are then compared with the experimental data, the modified SKM suggested by Rezaei and Knight (MSKM) and the SKM method. The study shows that there are good agreement between the results of the new analytical method and the experimental data.

With the expansion of the use of neural networks, increase deep of NN, and increase of network parameters, as well as the limitation of computational resources, the limitation of memory, and the incomprehensibility of these networks, the compression of neural networks is necessary. Compression must be intelligent, so as not to deprive us of the benefits of deep neural networks. Pruning is one of the compression methods that eliminate unnecessary network parameters. in recent research, Pruning has always been favored by researchers as far as a step called pruning at the initial design that pruned the initial network to include the benefits compression and pruning in the training and inference. this article reviews pruning techniques in deep neural networks with emphasis on Prune at initializing. First, the basics of pruning are discussed, then the types of pruning with the mathematical definition of each discussed, and finally, a more detailed study of pruning before network training has been done.

Sign Language is a visual language that uses hand shapes, facial expression gestures, and body language for communication. Linguistic studies have shown that Sign Languages, like spoken languages, are natural and meet the communication needs of their speakers. Sign Language is not a single language that can be universal, nor every country has a Sign Language equivalent to its own spoken language. There are many Sign Languages ​​in the world including the American, English, Japanese, Italian, Turkish, and Persian Sign Language. The Persian Sign Language is the natural language of the deaf in Iran. The ability to communicate effectively is an important step in establishing relationships and participation for the deaf in the community. Lack of support from the hearing community leads deaf people to isolation and other social problems. In order to overcome the obstacles that exist between hearing and hearing impaired people, a translator system is needed to translate messages into the Sign Language in accordance with their spoken language. A Sign Language translation system can be used by the deaf, the family and friends of such people, the blind-deaf people, the people who have lost their ability to speak due to some factors, the professors and teachers, the medical staff and also people who are interested in learning this language. This article, by providing an automatic translator system for translating from the Persian Language into the Persian Sign Language, aims to help the deaf in Iran to communicate better.

The proposed translator system requires the use of an architecture in order to convert the Persian text into the Persian Sign Language. Due to the nature of the Persian language as the input language of the translator system and also the problems inherent in the Persian Sign Language as the output language of the translator system, the architecture of the translation systems presented in other languages ​​cannot be presented and developed regarding the Persian language. One of the most important issues in the Persian Sign Language is lack of equivalent signs for some words in the Persian language. In order to overcome these problems, an architecture appropriate to the Persian language and the Persian Sign Language was presented. By using the proposed architecture, the proposed system, receives the Persian text as an input in the form of word or sentence from the user, and after performing the initial processing and by using the lexical transfer module, the system translates the text into the Persian Sign Language. Finally, by using an avatar, the Sign Language form corresponding the input of a word, phrase, or sentence can be represented. For this purpose, transfering the translated words into HamNoSys notation, turning the HamNoSys notation to SiGML language and then animating the 3D character with the help of this language are used.

To evaluate the translation system of the Persian language to the Persian Sign Language, a set of test sentences were used. After evaluating the system, it was found that the proposed system has acceptable efficiency, storage space and speed.

Research in interdisciplinary sciences is only effective and influential when research in all of the sciences involved is done equally and each science solves the challenges of its own field of study. For example, the most important challenge in completing the Persian Sign Language translation system is lack of linguistic research on Persian Sign Language. The proposed translation system is a combination of linguistic, social, and engineering sciences. Focusing on each of these areas and upgrading them will considerably improve the proposed system. However, the proposed system can improve the relationship between the hearing and the hearing impaired people to a great extent. One can focus on any of the proposed architecture modules and upgrade and improve each of them. It is also possible to integrate the section of emotions and facial expressions with the animated character so that the facial expressions of this avatar can change according to the conditions.  

Reducing transportation problems regarding safety, traffic and consequently the human, social and economic impacts is one of the main challenges of human societies. The incremental increase in driving violations and their harmful effects are one of the most important social issues in different countries. Multimedia/unstructured data has grown dramatically over the last few years. The share of big data and behavior detection in this regard are more likely to be addressed by controlling the mechanized violation of drivers. In the current system, control is carried out by human sources (police or operator). The purpose of this research is to control the mechanized violation of drivers based on distributed architecture and the Map / Reduce programming model that reduce processing time (CPU time) with Hadoop. In order to investigate and test the proposed system, information about cameras available in the city has been obtained from the Traffic Control Center of Yazd city in Iran. The results indicate that the processing time for a big data using Hadoop and executing the program only with a single slave-node decreases by more than 87% relative to the implementation of the program sequentially. Also, system performance increases by increasing the number of data nodes by more than 75%. The algorithm presented in this article without 24-hour operator is able to recognize inappropriate and violation behavior of drivers. Therefore, if the operator is not present, the driver's behavior will not be out of sight.

Early diagnosis of breast cancer plays a crucial role in treating the patient. Nowadays, data mining algorithms can provide intelligent methods in the health and treatment system that accurately detect breast cancer. The purpose of this study is breast cancer detection using ensemble classifier based on WBC and WDBC prepared databasesa. Our proposed model in the WBC database (reducing features by cfs+ optimizing samples using Resample+ ensemble classifier using data mining algorithms (kstar + random forest + Naïve Bayes and Bayes network)) has the best detection accuracy ( 100%), implementation time (0 seconds) and without any errors and on the WDBC database (reducing features by cfs+ optimizing samples using Resample+ ensemble classifier using data mining algorithms (IBK algorithm+ Naïve Bayes, Bayes network and kstar)) has an accuracy of 99/29, the implementation time is 0 seconds, and the mean absolute error is 0/007. The results of this study show that according to the ensemble classifier methods using data mining algorithms on the prepared database, new systems can be designed to help physicians that facilitate treatment processes.

The mainstream of society has isolated deaf people due to their inability to communicate with the deaf community. Currently, 1,400 deaf people are born in the country every year, most of whom have a hearing family. For this reason, parents in the early years of a child's life, due to unfamiliarity with sign language, cannot teach them a complete model of sign language. As the name implies, sign language uses gestures, or more precisely hand, eye, and standing postures to convey meaning. If, for example, the movement of the mouth or the general state of the face changes, the meaning of the word produced can also change, or in other words, another word is produced. The same principle applies to the movements of the fingers and hands. The most important factor through which human beings can form their relationships and connections is using different methods such as speaking, writing and sign language. The point that deaf or semi-deaf people are ignored in communities is because of their disability in forming intimate relationships, and also the lack of common training according to what is needed in the society. The text to sign language translation systems have been created to resolve communication problems among hearing and deaf persons. Due to unstructured Persian language as well as the problems of Persian sign language, using translators’ systems of other countries is not practicable. The present study seeks to design a system that receives Persian text as an input and changes it to Persian sign language after all the necessary natural language processing is done.

In doing this matter, several challenges are seen on our way:  a) The number of Persian sign language words which are few compared to the words in Persian language and b) The ambiguity and pun normally used in Persian language. So, similar systems’ architecture could not be applied. Therefore, a new architecture is proposed that uses the relation between the removal pun modules and input text invert module to the deaf world by using designed ontology.

Using this architecture, the efficiency of the Persian text to Persian sign language translator system has improved to 95% and increased the accuracy of the translator system by tripling.

Sign language interpreter system is designed to connect deaf people with other people in the community. In addition, the sign language translator system can be used to teach sign language. Implementing the Persian sign language translator system has challenges due to its unstructured as well as the problem of ambiguity of some words in Persian, as well as the problems in Persian sign language and the limitations in the number of known movements of this language. The paper addressed the challenges to an acceptable degree by presenting a new architecture. In the new architecture, two modules for recognizing and correcting puns and converting sentences into the world of the deaf have been used. Both modules use their own designed ontology.

Image processing is one of the most important issues in the field of artificial intelligence, which is used in various industrial, medical, military, and security systems. One of the most important applications of image processing is the extraction of different types of classification in the field of medical sciences. By using powerful algorithms in this field, intelligent systems can be invented that automatically understand and interpret the medical characteristics of individuals without the need to the physician supervision can discover useful information to help experts make good judgments. When the necessary parameters for the diagnosis of the disease increase, the diagnosis and prognosis of the disease becomes very difficult even for an expert, which is why computer diagnostic tools have been used in recent decades to help the physicians. This has led to a reduction in possible errors due to fatigue or inexperience of the specialist, and to provide the required medical data to the physician in less time and with more detail and accuracy. The purpose of this study is to improve the classification of new methods using a multi-layered model to address retinal diseases diagnosis.

This paper presents a multi-layer dictionary learning method for classification tasks.  Our multi-layer framework uses a label consistent in K-SVD algorithm to learn a discriminative dictionary for sparse coding in order to learn better features in retinal optical coherence tomography images. In addition to using class labels of training data, we also associate label information with each dictionary item (columns of the dictionary matrix) to enforce discrimination in sparse codes during dictionary learning process. In fact, it relies on a succession of sparse coding and pooling steps in order to find an effective representation of data for classification. Moreover, we apply Duke dataset for validating our algorithm: Duke spectral domain OCT (SD-OCT) dataset, consisting of volumetric scans acquired from 45 subjects 15 normal subjects, 15 AMD patients, and 15 DME patients.

Our classifier leads to a correct classification rate of 95.85% and 100.00% for normal and abnormal (DME and AMD). Experimental results demonstrate that our algorithm outperforms compared to many recent proposed supervised dictionary learning and sparse representation techniques. Conlusion: The results of this study were to provide an automatic system for the diagnosis of some retinal abnormalities in a way that it could do data analysis with high accuracy in comparison to other modern methods to diagnosis delicate patterns of OCT, separate images of normal and patient the normal and in two age-related macular degeneration diseases (AMD), and diabetic macular degeneration (DME), and help the physician to diagnose retinal pathology with great care. As a suggestion for professionals and future research, by generalizing this method to the more classes, we can cover the entire retinal myopia and use it as a potentially effective tool in computerized diagnosis and screening for retinal disease or in the wider eye area.  

With the development of technology, the use of wireless telecommunication networks for the various affairs is essential. These networks are one of the safest and most widely used networks, for instance, in medical care and remote patient monitoring. What matters is the quality of service in these networks. The purpose of this paper is to increase packet transduction in a wireless body area network by assuming that the use of time-based access protocol (TDMA) method is effective in increasing the accuracy of packets and increasing the quality of network service. Therefore, by presenting a protocol in the link layer of the Internet of Things, some of the parameters of the quality of service were improved. Compared with other methods, the proposed protocol yielded better results in terms of latency, energy consumption and throughput. The proposed algorithm improvement was 23.67% in the packet delay, 27.12% in the energy consumption, and 10% in throughput.

Due to the growth of digital images require efficient methods to annotate the images is sense. In this paper, a semi-supervised spectral clustering with relevance feedback is used to annotate digital photos which is overcome the local minima problem on clustering methods by using some labeled information given by users. Performance of the proposed method is tested on Corel 5K dataset and the results demonstrate the efficiency and accuracy of the proposed method compared with other clustering methods.

One of the most important stages in Character Recognition Systems is “Segmentation”, because any mistake will affect to all other tasks, especially to character recognition. This operation is more complex in Persian/Arabic writing than other Latin writing like English, and there has been an ongoing research on it. Other algorithms, that has been used as base as proposed algorithm, show 85% accuracy. In this paper, a new improved method has been presented by analyzing the visual features of the Persian/Arabic language. The proposed algorithm is able to segment existing fonts up to 98.5% accuracy or even 100% on some cases. The remaining error could be refined by applying a good character recognition technique and a precise vocabulary.