With the increasing desire of companies and organizations to employ interns in various situations, choosing the right person to participate in internships has become very important. Although the person who is selected for an internship must have relative knowledge and skills in the desired work fields,it should not be expert and experienced,because such people usually demand high wages. Community inquiry websites with many users can be used as one of the sources of intern knowledge. In previous research, statistical characteristics such as the number of answers, the number of specialized areas, the length of answers, and similar features have been proposed to identify potential interns,but the content of the user's answers has not been used to recognize the interns. This textual content is a rich resource for determining the breadth or depth of user knowledge and can be of great help in identifying potential trainees. In this research, a deep learning model called CNN-BiLSTM has been proposed to identify suitable people for internships based on the text of the answers they send to community inquiry websites. In addition, three machine learning models and four widely used deep learning models have also been used for comparison. Based on the obtained results, deep learning models have performed better in comparison with machine learning algorithms based on accuracy and F1 criteria. Also, among deep learning models, the proposed model has been able to show at least 7% higher accuracy and 2% higher F1 criterion than other models used to identify potential trainees.

Emotions are human mental states at a particular instance in time concerning one’s circumstances, mood, and relationships with others. Identifying emotions from the whispered speech is complicated as the conversation might be confidential. The representation of the speech relies on the magnitude of its information. Whispered speech is intelligible, a low-intensity signal, and varies from normal speech. Emotion identification is quite tricky from whispered speech. Both prosodic and spectral speech features help to identify emotions. The emotion identification in a whispered speech happens using prosodic speech features such as zero-crossing rate (ZCR), pitch, and spectral features that include spectral centroid, chroma STFT, Mel scale spectrogram, Mel-frequency cepstral coefficient (MFCC), Shifted Delta Cepstrum (SDC), and Spectral Flux. There are two parts to the proposed implementation. Bidirectional Long Short-Term Memory (BiLSTM) helps to identify the gender from the speech sample in the first step with SDC and pitch. The Deep Convolutional Neural Network (DCNN) model helps to identify the emotions in the second step. This implementation is evaluated with the help of wTIMIT data corpus and gives 98. 54% accuracy. Emotions have a dynamic effect on genders, so this implementation performs better than traditional approaches. This approach helps to design online learning management systems, different applications for mobile devices, checking cyber-criminal activities, emotion detection for older people, automatic speaker identification and authentication, forensics, and surveillance

The proliferation of fake news, characterized by the dissemination of inaccurate information to deceive audiences, has become a pressing concern in recent times. Traditional approaches to phony news detection, often focused on analyzing Twitter content, are susceptible to noise and variations in input sequences, leading to suboptimal performance. To address these challenges, this study proposes a novel method called Multi-Head Attention-Hierarchical Bidirectional Long Short-Term Memory (MHA-HBiLSTM) Networks. Our approach involves two phases: training and testing, wherein we employ tweet pre-processing techniques such as stemming, punctuation removal, stop-word elimination, URL handling, and Twitter control removal. Features are represented using the Glove word embedding technique for experimental evaluation and comparison. The MHA-HBiLSTM model integrates multi-head attention and hierarchical concepts, allowing meaningful information extraction from Twitter data. Notably, our model utilizes dual-level attention mechanisms and a hierarchical structure, reflecting the inherent hierarchy in documents and prioritizing key material during document representation. The effectiveness of the proposed MHA-HBiLSTM algorithm is evaluated using the Whale & Multi-Verse (W-MVO) Optimizer approach, with tests conducted on Kaggle and FakeNewsNet datasets. Comparative analysis with traditional machine learning approaches and deep learning models demonstrates the superior performance of the MHA-HBiLSTM approach in fake news detection.

Solar energy forecasting is necessary due to its variable and fluctuating nature, but it is also a challenge to predict accurately behaviour of solar irradiation. To capture this, the proposed methodology uses an ensemble model combined with error minimization and CEEMDAN Pre-processing technique. In this paper, data of two locations are used to predict short term forecasting of solar irradiation using seven developed models based on the proposed procedure. The use of hourly forecasting, CEEMDAN method, error minimization and ensemble hybrid model enhance the anti-interference capability of all developed model. Four-year data of New Delhi and Ahmedabad is used and sourced from NSRDB website. Out of all the proposed models CEEMDAN-CNN-BiLSTM-MLP with CEEMDAN_IMF_18 configured signal processing approach achieved least average RMSE, n-RMSE and MAE of both locations with values 13.215 W/m2, 7.13% and 8.605 W/m2 respectively and have maximum average R2 (99.205%). When compared to persistence model, proposed model with this configuration was able to outperform with average percentage improvement 87.63%, 86.78%, 87.17% and 17.875% in terms of  , ,  and   respectively. The proposed model outperforms existing techniques for solar irradiation forecasting, demonstrating greater efficiency and reliability, making it a valuable reference for future performance optimization.

Today, with the significant growth of artificial intelligence and its products, many opportunities and threats have emerged. One of the most famous and popular products of artificial intelligence is text generation, also called machine text. In this research, a new method is introduced that combines features extracted from the text with its structural features, thus creating an automatic discriminator to distinguish between human-written text and artificial intelligence-generated text. The introduced method consists of two parts, the first part: the extended BERT (RoBERTa) model and the bidirectional long-term short-term memory (BiLSTM) model, which are improved with the fusion layer. The second part: the structural features of the text are extracted using a writing style-based method. Finally, the output of the model parts is combined together, and in this way, the model distinguishes human-written text from machine-generated text. The results of this research show that the proposed method is capable of recognizing machine texts with 90% accuracy and exhibits good performance.

The retail market industry is one of the industries that affects the economies of countries, the life of which depends on the level of satisfaction and trust of customers to buy from these markets. In such a situation, the retail market industry is trying to provide conditions for customer feedback and interaction with retailers based on web pages and online platforms. Because the analysis of published opinions play a role not only in determining customer satisfaction but also in improving products. Therefore, in recent years, sentiment analysis techniques in order to analyze and summarize opinions, has been considered by researchers in various fields, especially the retail market industry.

Background: The rise of smartphone sensors, especially accelerometers, has expanded the scope of Human Activity Recognition (HAR). HAR plays a key role in monitoring student health by offering real-time insights into physical activity and promoting healthier behaviors. Objective: The aim of this study is to develop an optimized deep learning model to monitor and classify student activities, using accelerometer data for real-time health monitoring. Methods: This study developed and optimized a novel deep learning framework using modified version of Bidirectional Long Short-Term Memory (BiLSTM) networks, enhanced by the Grey Wolf Optimizer (GWO). The BiLSTM framework automates the feature learning process from raw accelerometer data, while GWO optimizes the hyperparameters to improve sequence processing and overall model performance. We employed public datasets, UCI-HAR and WISDM, for validation, using cross-validation to ensure model robustness. The edge computing approach was implemented to enable real-time processing. Results: The proposed BiLSTM-GWO framework achieved a classification accuracy of 97. 68%, outperforming existing methods in recognizing student activities. The model showed enhanced performance in distinguishing between activities such as walking, sitting, and stair climbing, significantly reducing misclassification errors. In addition to accuracy, metrics such as precision, recall, and F1 score were evaluated, all showing improvement. GWO optimization also accelerated convergence, enhancing suitability for real-time applications. Conclusions: The integration of edge computing into the framework provides real-time analysis and resource efficiency, making it highly suitable for health monitoring applications in educational settings.

Sentiment analysis of online doctor reviews helps patients to better evaluate and select the related doctors based on the previous patients' satisfaction. Although some studies are addressing this problem in the English language, only one preliminary study has been reported for the Persian language. In this study, we propose a new evolutionary deep model for sentiment analysis of Persian online doctor reviews. The proposed method utilizes both Persian reviews and their English translations as inputs of two separate deep models. Then, the outputs of the two models are combined in a single vector which is used for deciding the sentiment polarity of the review in the last layer of the proposed deep model. To improve the performance of the system, we propose an evolutionary approach to optimize the hyperparameters of the proposed deep model. We also compared three evolutionary algorithms, namely, Ant Colony Optimization (ACO), Genetic Algorithm (GA), and Gray Wolf Optimization (GWO) algorithm, for this purpose. We evaluated the proposed model in two phases,In the first phase, we compared four deep models, namely, long shortterm memory (LSTM), convolutional neural network (CNN), a hybrid of LSTM and CNN, and a bidirectional LSTM (BiLSTM) model with four traditional machine learning models including Naïve Bayes (NB), decision tree (DT), support vector machines (SVM), and random forest (RF). The results showed that the BiLSTM and CNN models outperform other methods, significantly. In the second phase, we compared the optimized version of two proposed bi-lingual models in which either two BiLSTM or two CNN models were used in parallel for processing Persian and English reviews. The results indicated that the optimization of the CNN using ACO and the optimization of BiLSTM using a genetic algorithm can achieve the best performance among other combinations of two deep models and three optimization algorithms. In the current study, we proposed two deep models for bi-lingual sentiment analysis of online Persian doctor reviews. Moreover, we optimized the proposed models using ACO, genetic algorithm, and gray wolf optimization methods. The results indicated that the proposed bi-lingual model outperforms a similar model using only Persian or English reviews. Also, optimizing the parameter of proposed deep models using ACO or genetic algorithms improved the performance of the models.

Introduction: Neurodegenerative diseases (NDDs) present substantial challenges due to their impact on movement, emphasizing the critical role of biomedical engineering research in clinical diagnosis. Measuring the biomechanical properties of gait during walking can provide valuable insights into the movement pattern of NDDs and has great promise for developing non-invasive automated NDD classification techniques.  Methods: Based on the GaitNDD database, two experimental trials were conducted on healthy controls (HCs) and three NDDs: Parkinson disease (PD), amyotrophic lateral sclerosis (ALS), and Huntington disease (HD), showcasing a comprehensive analysis of 1-dimensional and 2-dimensional force gait features. In the first trial, two time-frequency feature sequences were extracted from right, left, and combined feet during a walking task, feeding a bidirectional long short-term memory (BiLSTM) network. The second trial involves constructing spectrogram images of the gait signal as input for 3 popular pre-trained convolutional neural networks (CNNs): AlexNet, GoogLeNet, and VGG16.  Results: VGG16 emerges as the standout performer, achieving a remarkable accuracy of 99.91%, sensitivity of 99.93%, and specificity of 99.97% for automatic 4-class NDD detection using high-level features from the right foot gait signal. BiLSTM performance significantly improved when fed with VGG16-extracted high-level features, surpassing hand-crafted features. Conclusion: The study underscores the superiority of CNNs, particularly VGG16, in extracting high-level features from spectrogram-derived vertical ground reaction force (vGRF) signals for robust NDD classification. The hybrid VGG16-BiLSTM approach demonstrates enhanced performance, affirming the synergistic benefits of combining deep learning techniques. Overall, the CNN high-level features derived from vGRF signal spectrograms provide valuable insights into NDD groups, offering a promising avenue for understanding diverse mechanisms underlying gait-related conditions.