After an extended outage in a power system, restoration of feeders with high penetration levels of thermostatically controlled loads (TCLs) will be a challenging problem. By restoring these feeders, the demand will increase several times the normal conditions which is called the cold load pick-up (CLPU) phenomenon. This can lead to overload of equipment such as power transformers in transmission substations. In order to increase the reliability of network load supply, in transmission substations, usually several transformers are used which can be operated in parallel by closing circuit breakers on the secondary side called bus section (BS). When the restoration process is long and the loads are cold, more loads can be restored by closing the BSs and paralleling the transformers. On the other hand, it is not possible to close all the BSs simultaneously because paralleling all the transformers at the same time increases the short circuit level on the secondary side. Furthermore, in power systems, some loads are voltage-dependent and by reducing the voltage, their demand can be decreased and therefore the restoration process is expedited. Due to the increase in demand caused by CLPU, additional stress is applied to the power system, which can lead to operational constraints violation and even long-term voltage instability. Therefore, at the time of restoration of each feeder, the settings of voltage control devices are updated to avoid the violation of operational constraints and the occurrence of wider events. In this paper, in order to reduce cold load restoration time, closing the bus sections and voltage reduction are used simultaneously. Also, short circuit and operational constraints are considered. The cold load restoration problem is formulated as a mixed integer nonlinear programming and is solved using the pattern search algorithm. Since there are many decision variables and solving the problem is time-consuming, a combination of deep learning networks, pattern search algorithm and parallel computing is used to accelerate the computations.

Lack of awareness and knowledge of microservices-specific security challenges and solutions often leads to ill-informed security decisions in microservices system development. We claim that identifying and leveraging security discussions scattered in existing microservices systems can partially close this gap. We define security discussion as "a paragraph from developer discussions that includes design decisions, challenges, or solutions relating to security". We first surveyed 67 practitioners and found that securing microservices systems is a unique challenge and that having access to security discussions is useful for making security decisions. The survey also confirms the usefulness of potential tools that can automatically identify such security discussions. We developed fifteen machine/deep learning models to automatically identify security discussions. We applied these models on a manually constructed dataset consisting of 4,813 security discussions and 12,464 non-security discussions. We found that all the models can effectively identify security discussions: an average precision of 84.86%, recall of 72.80%, F1-score of 77.89%, AUC of 83.75% and G-mean 82.77%. DeepM1, a deep learning model, performs the best, achieving above 84% in all metrics and significantly outperforms three baselines. Finally, the practitioners' feedback collected from a validation survey reveals that security discussions identified by DeepM1 have promising applications in practice.

Instagram is a popular social media that has a wide range of active users from ordinary people to artists and official users. Instagram posts are widely used by users to share text, image or video. Many users use text to describe or complement the images they share. To analyze the sentiment of such posts, both the content of the text and the image should be considered at the same time. This requires modelling of the relationship between the text and image modalities. To address this problem, we propose a multimodal deep learning method. The proposed method utilizes a bi-directional gated recurrent unit (bi-GRU) for processing text comments and a 2-dimensional convolutional neural network (2CNN) for analyzing images. In order to assess the performance of the proposed model, we introduce a new dataset of Instagram posts, MPerInst, containing 512 pairs of images and their corresponding comments written in the Persian language. Implementation results shows that employing both text and image modalities improves polarity detection accuracy and F1-scrore by 23% and 0.24 compared to using only image and text modalities, respectively. Moreover, the proposed model outperforms 11 similar deep fusion models by 11% and 0.1 in terms of accuracy and F1-score. Both the dataset and the codes of our proposed model are publicly available for probable future use.

Undoubtedly, coronavirus (COVID-19) has caused one of the biggest challenges of all times. The ongoing COVID-19 pandemic has caused more than 150 million infected cases and one million deaths globally as of May 5, 2021. Understanding the sentiment of people expressed in their social media comments can help in monitoring, controlling, and ultimately eradicating the disease. This is a sensitive matter as the threat of infectious disease significantly affects the way people think and behave in various ways. In this study, we proposed a novel method based on the fusion of four deep learning and one classical supervised machine learning model for sentiment analysis of coronavirus-related tweets from eight countries. Also, we analyzed coronavirus-related searches using Google Trends to better understand the change in the sentiment pattern at different times and places. Our findings reveal that the coronavirus attracted the attention of people from different countries at different times in varying intensities. Also, the sentiment in their tweets is correlated to the news and events that occurred in their countries including the number of newly infected cases, number of recoveries and deaths. Moreover, common sentiment patterns can be observed in various countries during the spread of the virus. We believe that different social media platforms have great impact on raising people's awareness about the importance of this disease as well as promoting preventive measures among people in the community.

Sentiment analysis has been a hot research topic in natural language processing and data mining fields in the last decade. Recently, deep neural network (DNN) models are being applied to sentiment analysis tasks to obtain promising results. Among various neural architectures applied for sentiment analysis, long short-term memory (LSTM) models and its variants such as gated recurrent unit (GRU) have attracted increasing attention. Although these models are capable of processing sequences of arbitrary length, using them in the feature extraction layer of a DNN makes the feature space high dimensional. Another drawback of such models is that they consider different features equally important. To address these problems, we propose an Attention-based Bidirectional CNN-RNN Deep Model (ABCDM). By utilizing two independent bidirectional LSTM and GRU layers, ABCDM will extract both past and future contexts by considering temporal information flow in both directions. Also, the attention mechanism is applied on the outputs of bidirectional layers of ABCDM to put more or less emphasis on different words. To reduce the dimensionality of features and extract position-invariant local features, ABCDM utilizes convolution and pooling mechanisms. The effectiveness of ABCDM is evaluated on sentiment polarity detection which is the most common and essential task of sentiment analysis. Experiments were conducted on five review and three Twitter datasets. The results of comparing ABCDM with six recently proposed DNNs for sentiment analysis show that ABCDM achieves state-of-the-art results on both long review and short tweet polarity classification.

With the significant expansion of Q&A forums and the increasing need for users to access useful information, identifying quality content in text forums is of particular importance. Previous studies have focused on extracting several types of quality features from text that may be a time and labor-intensive task. To address this problem, in this paper, a long short-term memory (LSTM) deep neural network model is proposed to determine high-quality responses of users in text forums using only raw text of user replies. In the proposed model, embeddings from language models (ELMo) are usesd to represent words in vectors or embeddings. The proposed model is evaluated on two datasets: The TripAdvisor for New York City (NYC) and the Ubuntu Linux distribution online forums. Comparison of the results obtained using the proposed model and support vector machines (SVM), linear regression (LR), artificial neural networks (ANN), and naïve Bayes (NB) algorithms showed that, using only textual features, the accuracy of the proposed model was 43% and 28% higher compared to the highest accuracy obtained by the four traditional machine learning (ML) algorithms on the NYC and the Ubuntu datasets, respectively. This improvement was about 17% and 16% compared to the best results obtained by ML algorithms using both textual and quality dimension features.

Everyday many online product sales websites and specialized reviewing forums publish a massive volume of human-generated product reviews. People use these reviews as valuable free source of knowledge when decide to buy products. Therefore, an accurate automated system for distinguishing useful reviews from non-useful ones is of great importance. This article presents a new model for specifying the usefulness of comments using the textual features extracted from the reviews. Various types of features including emotion-related, linguistic and text-related features, valence, arousal, and dominance (VAD) values, review-length and polarity of comments are exploited in this study. Moreover, two new algorithms are presented: an improved evidential algorithm for emotion recognition, and an algorithm for extracting VAD values for each review. Finally, the usefulness of reviews is predicted using the mentioned features and an improved Dempster–Shafer score fusion algorithm. The proposed method is applied to review datasets of Books and Video Games of Amazon. The results show that combining the features associated with emotions, features of VAD, and text-related features improves the accuracy of predicting the usefulness of reviews. Also, in comparison with the original Dempster–Shafer method, the precision of the improved Dempster–Shafer algorithm for both datasets is 15% and 11% higher, respectively.

Online medical reviews contain patients' subjective evaluations and reflect their satisfaction with the treatment process and doctors. Mining and analysis of sentiment expressed in these medical data may be vital for different applications including adverse drug effects detection, doctor recommendation, and healthcare quality assessment. Nevertheless, medical sentiment analysis is a challenging and complex task because patients who write the reviews are usually non-professional users and tend to use informal language. The problem is more challenging in the Persian language due to its resource scarcity and complex structure. In this study, we introduce PODOR, a Persian dataset of online doctor reviews extracted from social web. Also, we propose a deep model based on the bidirectional long short-term memory for polarity detection of PODOR reviews. To show the effectiveness and suitability of the proposed model, we compared the model with six traditional supervised machine learning methods and three deep models. Preliminary comparative results indicated that our model outperformed traditional methods by 8% and 7%, and deep models by 2% and 3% in terms of accuracy and f1-measure.

Opinion mining is a subfield of data mining and natural language processing that concerns with extracting users’ opinion and attitude towards products or services from their comments on the Web. Persian opinion mining, in contrast to its counterpart in English, is a totally new field of study and hence, it has not received the attention it deserves. Existing methods for opinion mining in the Persian language may be classified into machine learning– and lexicon-based approaches. These methods have been proposed and successfully used for polarity-detection problem. However, when they should be used for more complex tasks like rating prediction, their results are not desirable. In this study, first an exhaustive investigation of machine learning– and lexicon-based methods is performed. Then, a new hybrid method is proposed for rating-prediction problem in the Persian language. Finally, the effect of machine learning component, feature-selection method, normalisation method and combination level are investigated. The experimental results on a large data set containing 16,000 Persian customers’ review show that this proposed system achieves higher performance in comparison to Naïve Bayes algorithm and a pure lexicon-based method. Moreover, results demonstrate that this proposed method may also be successfully used for polarity detection.

Coronary artery disease (CAD) is a leading cause of death worldwide and is associated with high health-care expenditure. Researchers are motivated to apply machine learning (ML) for quick and accurate detection of CAD. The performance of the automated systems depends on the quality of features used. Clinical CAD datasets contain different features with varying degrees of association with CAD. To extract such features, we developed a novel hybrid feature selection algorithm called heterogeneous hybrid feature selection (2HFS). In this work, we used Nasarian CAD dataset, in which work place and environmental features are also considered, in addition to other clinical features. Synthetic minority over-sampling technique (SMOTE) and Adaptive synthetic (ADASYN) are used to handle the imbalance in the dataset. Decision tree (DT), Gaussian Naive Bayes (GNB), Random Forest (RF), and XGBoost classifiers are used. 2HFS-selected features are then input into these classifier algorithms. Our results show that, the proposed feature selection method has yielded the classification accuracy of 81.23% with SMOTE and XGBoost classifier. We have also tested our approach with other well-known CAD datasets: Hungarian dataset, Long-beach-va dataset, and Z-Alizadeh Sani dataset. We have obtained 83.94%, 81.58% and 92.58% for Hungarian dataset, Long-beach-va dataset, and Z-Alizadeh Sani dataset, respectively. Hence, our experimental results confirm the effectiveness of our proposed feature selection algorithm as compared to the existing state-of-the-art techniques which yielded outstanding results for the development of automated CAD systems. (C) 2020 Elsevier B.V. All rights reserved.

More than 50 million journal papers will have been published by the end of 2019 with 2 million more journal papers published every year. The number of conference papers is even higher, and millions of other types of scientific research are added to the knowledge base every year. Scientific databases such as Web of Science, Scopus, and PubMed index millions of scientific papers and Google Scholar indexes a huge amount of scientific knowledge across diverse domains. However, current systems provide long lists of results when users attempt to find relevant papers, leaving them with little choice other than manually skimming through the lists. This article surveys different techniques used to identify relevant research papers by knowledge-based organizations. We categorized current literature content as content, metadata, collaborative filtering, and citation based techniques and identified the strengths and limitation for each approach. Further, we evaluated the published techniques and research-based products used to identify relevant documents and identified the strengths and limitations of each approach. This research will greatly help to understand current state-of-the-art techniques internal workings for finding relevant papers, understand the relevant strengths and limitations, and explore previously proposed techniques targeting this area.

Alaska is at the forefront of climate change and subject to salient challenges including energy consumption. It is important to understand Alaskans' perceptions and opinions about energy consumption to solve Alaska's domestic energy problems and creating a sustainable future. However, it is challenging to collect public opinions about energy consumption using conventional survey methods, which are often expensive, labor-intensive, and slow. This study utilizes information-rich Twitter data to investigate Alaskans' perceptions and opinions on various energy sources and in particular clean energy sources. Using the geotagged Twitter data collected in Alaska from 2014 to 2016, a lexicon-based sentiment analysis approach was first applied to analyze the polarity in the expressed opinions. Further, a novel fuzzy-based theory is employed to derive the sentiment of the opinion in each tweet. The results indicate that there is a valuable growth rate for a set of energy-related keywords, such as "sun", "power", and "nuclear". The rank of top 20 renewable energy-related keywords shows the word "Tidal" has the highest ranking followed by "solar panel". Moreover, the attention to various types of energy is increasing dramatically among Alaskans. Importantly, Alaskans' attitudes toward energy and renewable energy changed positively from 2014 to 2016, indicating that Alaskans' energy choices are more acceptive towards or even favor renewable energy in the future.

Sentiment analysis is one of the natural language processing tasks used to find reviews expressed in online texts and classify them into different classes. One of the most important factors affecting the efficiency of sentiment analysis methods is the aggregation algorithm used for scores combination. Recently, Dempster–Shafer algorithm has been used for scores aggregation. This algorithm has a higher precision than common methods such as average, weighed average, product and voting, but the problem with this algorithm is the aggregation of a dominant high or low score that is always selected by the algorithm as the overall score. In the current research, a new method is proposed for scores aggregation that employs both the most and the second probable classes to predict the final score. The proposed approach considers every review as a set of sentences each of which has its own sentiment orientation and score and computes the probability of belonging of every sentence to different classes in a five-star scale using a pure lexicon-based system. These probabilities are then used for document-level sentiment detection. To this aim, two-point structure is used to improve the Dempster–Shafer aggregation algorithm. The proposed method is applied to review datasets of TripAdvisor and CitySearch which have been used in previous studies. The obtained results show that in comparison with the original Dempster–Shafer aggregation method, the precision of the proposed method for both datasets is 23% and 27% higher, respectively.

Product reviews are one of the most important types of user-generated contents that are becoming more and more available. These reviews are valuable sources of knowledge for users who want to make purchasing decisions and for producers who want to improve their products and services. However, not all product reviews are equally helpful and this makes the process of finding helpful reviews among the massive number of similar reviews very challenging. To address this problem, automatic review helpfulness prediction systems are designed to classify reviews according to their content. In this study, a deep model is proposed to utilize content-based, semantic, sentiment, and metadata features of reviews for predicting review helpfulness. In the proposed method, convolution layer is used for learning feature maps and gated recurrent units are employed for exploiting sequential context. The results of comparing the proposed method with five traditional learning methods and two deep models trained on the same types of features shows that the proposed method outperforms other methods by 4% and 2% in terms of F1-measure and accuracy. Moreover, results reveal that both textual and metadata features are important in detecting helpful reviews. The findings of this study may help online retailers to efficiently rank the product reviews.

: In the past decade, the study of human opinions, feelings and tendencies has been very effective in the decision-making of managers and individuals. Machine learning algorithms play an important role in the field of opinion mining, but they suffer from a big problem: most of the machine learning algorithms assume that the feature dimensions and data distribution are equal, but most of real-world applications don't follow these assumptions. In fact, the data that the algorithm will receive in the future may have different dimensions or distributions. In this article, a new method for improving sentiment analysis of opinions is proposed by the aid of feature-based transfer learning. In the proposed method, initially, the feature or topic of the opinion in the source language domain is identified. Then, by collecting adjectives, adverbs and totally a package of probabilities about that feature and by translating it into the target language, learning from the source language is transferred into the target language. An analysis of the proposed method on the data available at the Amazon store as the source domain indicates that by creating a pattern of feature transferring in English, the polarity of 77% of the opinions in Persian (recorded at the Digikala store) can be extracted that outperforms the SCL, SFA and TCA models with 9, 5 and 5 percent respectively.

Classification of noisy data has been a longstanding topic in data mining and machine learning. Many scientists have proposed effective methods to detect and eliminate such data in diverse real-world datasets. In this paper, we deal with mislabeled instances in supervised learning, including majority voting filtering and consensus voting filtering. The majority voting procedure usually incorrectly identifies many correct instances as noisy, whereas the consensus voting procedure is not able to detect at all many noisy instances. Our new method minimizes the majority and consensus filtering weaknesses by providing a novel class noise detection strategy, namely a high agreement voting filtering with mixed strategy, which proceeds by removing strong and semi-strong noisy records from the dataset as well as by relabeling weak noisy data. The proposed method, designed for binary classification problems, outperforms the high agreement voting filtering procedure. Extensive experiments conducted with 16 real datasets, using four noise filtering methods with two levels of class noise (10% and 15%), prove the superiority of the proposed methodology.

Nowadays, the development of new computer-based technologies has led to rapid increase in the volume of user-generated textual content on the website. Patient-written medical and health-care reviews are among the most valuable and useful textual content on social media which have not been studied extensively by researchers in the fields of natural language processing (NLP) and data mining. These reviews offer insights into the interaction of patients with doctors, treatment, and their satisfaction or frustration with the delivery of healthcare services. In this study, we propose two deep fusion models based on three-way decision theory to analyze the drug reviews. The first fusion model, 3-way fusion of one deep model with a traditional learning algorithm (3W1DT) developed using a deep learning method as a primary classifier and a traditional learning method as the secondary method that is used when the confidence of the deep method during classification of test samples is low. In the second proposed deep fusion model, 3-way fusion of three deep models with a traditional model (3W3DT), three deep and one traditional models are trained on the entire training data and each classifies the test sample individually. Then, the most confident classifier is selected to classify the test drug review. Our results on the reviews based on Drugs.com dataset show that both proposed 3W1DT and 3W3DT methods outperformed the traditional and deep learning methods by 4% and the 3W3DT outperformed 3W1DT by 2% in terms of accuracy and F1-measure.

Online shopping websites like Amazon stipulate a platform to the users where they can share their opinions about different products. Recently, it has been identified that prior to the purchasing, 81% of the users explore different online platforms in order to assess the reliability of product that they intend to buy. The reviews of different users are expressed by using natural language, which help a user to make an informed decision. From past few years, scientific community has payed attention to automatically specify the meaning of review through Sentiment Analysis. Sentiment Analysis is a research area which is gradually being evolved thus, helping the users to tackle the sentiment hidden in a review. To date, different sentiment analysis-based studies have been conducted in literature. For sentiment classification, the core ingredient is the exploitation of polarity bearing words present in the reviews e.g. adjectives, verbs, and adverbs etc. Different studies suggest the importance of different forms of adverbs in sentiment classification task. In literature, it has been reported that general adverbs strongly help to classify sentiments with better accuracy whereas other suggest that degree adverbs are important for sentiment classification. There are ten distinct forms of adverbs such as general adverbs, general superlative adverbs, general comparative adverbs, general-wh adverbs, degree adverbs, degree superlative adverbs, degree comparative adverbs, degree-wh adverbs, time adverbs and locative adverbs. In this paper, we intend to tackle a question that what is the impact of different forms of adverb on the classification of sentiments? For this, the impacts of all these forms have been evaluated on 51,005 reviews of two products, office products and musical DVDs acquired from Amazon. The outcomes of study revealed that two general superlative adverbs and degree-wh adverb hold more impact than the other forms of adverbs. The general superlative adverbs have attained F-measure of 0.86 and degree-wh adverbs have attained F-measure of 0.80.