Processing Model Research Articles

Оценка альтернатив: новые подходы принятия решений в условиях определенности

The decision-making process plays an important role in the development of society in general and the effective management of organizations in particular. In today's world, where the complexity of tasks and the competitive environment are constantly growing, making the right and informed decisions becomes a key element of success. Any person, group of people or organization is constantly faced with the need to choose between several possible courses of action, called alternatives, whether in the personal sphere or in the work of organizations. To effectively make complex decisions within the activities of organizations, models for making optimal decisions have been developed. The most popular of them is the model of decision-making under conditions of certainty, when the outcome of the decision is not influenced by random factors. All this confirms the relevance of scientific research in the field of decision making. Aim. The purpose of the study is to develop a methodology for assessing the degree of attractiveness of alternatives based on objective quantitative or subjective qualitative criteria, which will improve the quality of decisions made. The methodology is based on mathematical models for processing information containing the characteristics of alternatives according to evaluation criteria. Materials and methods. The methodology for evaluating alternatives presented in the work is based on mathematical models of information processing, which are based on methods of mathematical modeling, optimization methods, linear programming, theory of fuzzy sets, theory of latent variables, as well as methods of decision-making under conditions of uncertainty and methods of expert assessment. Results. The relevance of the development of a model for evalua¬ting construction projects according to quantitative criteria is shown, ways of solving the tasks are given and justified. A mathematical model has been developed that makes it possible to evaluate construction projects by objective indicators based on the theory of fuzzy sets and the Rush model for estimating latent variables. On the basis of numerical methods, the adequacy of the received estimates of construction projects is substantiated, the methodology for implementing computational procedures necessary for the practical implementation of evaluation models is described. Conclusion. The application of the given methodology in the decision-making procedure under conditions of certainty will improve the quality of decisions made and reduce the risks of their implementation.

Information retrieval and classification of real-time multi-source hurricane evacuation notices

In the United States, tracking time-sensitive critical information such as evacuation notices for approaching natural hazards like hurricanes poses a significant challenge. This difficulty arises from the rapid issuing and distribution of these notices by numerous local authorities that may span multiple states. Additionally, evacuation notices often undergo frequent updates, and they are distributed through various online portals lacking standard formats. Unlike weather warnings, there is no centralized database for live (or past) evacuation notices. To meet the demand of data for crisis-related government notices, we developed an approach to detect and retrieve locally issued hurricane evacuation notices in real time. Evacuation-related text data were collected mainly through spatially targeted information retrieval. These data were then manually labeled and used to train natural language processing (NLP) models. The NLP models in turn classified all text data into one of the three notice categories, i.e., mandatory evacuation notices, voluntary evacuation notices, and not an evacuation notice. The classification of mandatory evacuation notices achieved a very high accuracy (recall = 96 %). These NLP models, when applied to future hurricanes, will provide real-time evacuation notices for situation awareness to higher-level government agencies and news outlets. The archived evacuation notices serve as a valuable resource for scholars to study government responses to weather warnings and individual behaviors influenced by evacuation history. Our research framework may extend to other disaster types, allowing for rapid and targeted retrieval, classification, redistribution, and archiving of real-time government orders and notifications.

Optimization of mechanical and safety properties by designing interface characteristics within energetic composites

The interfacial structure has an important effect on the mechanical properties and safety of the energetic material. In this work, a mesostructure model reflecting the real internal structure of PBX is established through image digital modeling and vectorization processing technology. The microscopic molecular structure model of PBX is constructed by molecular dynamics, and the interface bonding energy is calculated and transferred to the mesostructure model. Numerical simulations are used to study the influence of the interface roughness on the dynamic compression and impact ignition response of PBX, and to regulate and optimize the mechanical properties and safety of the explosive to obtain the optimal design of the surface roughness of the explosive crystal. The results show that the critical hot spot density of PBX ignition under impact loading is 0.68 mm−2. The improvement of crystal surface roughness can improve the mechanical properties of materials, but at the same time it can improve the impact ignition sensitivity and reduce the safety of materials. The optimal friction coefficient range for the crystal surface that satisfies both the mechanical properties and safety of PBX is 0.06–0.12. This work can provide a reference basis for the formulation design and production processing of energetic materials.

Artificial intelligence for hearing loss prevention, diagnosis, and management

This paper explores the transformative impact of artificial intelligence (AI), particularly machine learning (ML), on diagnosing and treating hearing loss, which affects over 5% of the global population across all ages and demographics. AI encompasses various applications, from natural language processing models like ChatGPT to image recognition systems; however, this paper focuses on ML, a subfield of AI that can revolutionize audiology by enhancing early detection, formulating personalized rehabilitation plans, and integrating electronic health records for streamlined patient care. The integration of ML into audiometry, termed "computational audiology," allows for automated, accurate hearing tests. AI algorithms can process vast data sets, provide detailed audiograms, and facilitate early detection of hearing impairments. Research shows ML's effectiveness in classifying audiograms, conducting automated audiometry, and predicting hearing loss based on noise exposure and genetics. These advancements suggest that AI can make audiological diagnostics and treatment more accessible and efficient. The future of audiology lies in the seamless integration of AI technologies. Collaborative efforts between audiologists, AI experts, and individuals with hearing loss are essential to overcome challenges and leverage AI's full potential. Continued research and development will enhance AI applications in audiology, improving patient outcomes and quality of life worldwide.

ChatGPT, Bard, and Bing Chat are large language processing models that answered OITE questions with a similar accuracy to first-year orthopaedic surgery residents.

PurposeTo assess ChatGPT, Bard, and BingChat’s ability to generate accurate orthopaedic diagnosis or corresponding treatments by comparing their performance on the Orthopaedic In-Training Examination (OITE) to orthopaedic trainees. MethodsOITE question sets from 2021 and 2022 were compiled to form a large set of 420 questions. ChatGPT (GPT3.5), Bard, and BingChat were instructed to select one of the provided responses to each question. The accuracy of composite questions was recorded and comparatively analyzed to human cohorts including medical students and orthopaedic residents, stratified by post-graduate year. ResultsChatGPT correctly answered 46.3% of composite questions whereas BingChat correctly answered 52.4% and Bard correctly answered 51.4% of questions on the OITE. Upon excluding image-associated questions, ChatGPT, BingChat, and Bard’s overall accuracies improved to 49.1%, 53.5%, and 56.8%, respectively. Medical students and orthopaedic residents (PGY1-5) correctly answered 30.8%, 53.1%, 60.4%, 66.6%, 70.0%, and 71.9%, respectively. ConclusionChatGPT, Bard, and BingChat are AI models that answered OITE questions with an accuracy similar to that of first-year orthopaedic surgery residents. ChatGPT, Bard, and BingChat achieved this result without using images or other supplementary media that human test takers are provided.

Community-engaged artificial intelligence research: A scoping review.

The degree to which artificial intelligence healthcare research is informed by data and stakeholders from community settings has not been previously described. As communities are the principal location of healthcare delivery, engaging them could represent an important opportunity to improve scientific quality. This scoping review systematically maps what is known and unknown about community-engaged artificial intelligence research and identifies opportunities to optimize the generalizability of these applications through involvement of community stakeholders and data throughout model development, validation, and implementation. Embase, PubMed, and MEDLINE databases were searched for articles describing artificial intelligence or machine learning healthcare applications with community involvement in model development, validation, or implementation. Model architecture and performance, the nature of community engagement, and barriers or facilitators to community engagement were reported according to PRISMA extension for Scoping Reviews guidelines. Of approximately 10,880 articles describing artificial intelligence healthcare applications, 21 (0.2%) described community involvement. All articles derived data from community settings, most commonly by leveraging existing datasets and sources that included community subjects, and often bolstered by internet-based data acquisition and subject recruitment. Only one article described inclusion of community stakeholders in designing an application-a natural language processing model that detected cases of likely child abuse with 90% accuracy using harmonized electronic health record notes from both hospital and community practice settings. The primary barrier to including community-derived data was small sample sizes, which may have affected 11 of the 21 studies (53%), introducing substantial risk for overfitting that threatens generalizability. Community engagement in artificial intelligence healthcare application development, validation, or implementation is rare. As healthcare delivery occurs primarily in community settings, investigators should consider engaging community stakeholders in user-centered design, usability, and clinical implementation studies to optimize generalizability.

Processing calibration data of low-temperature thermometer based on clustering algorithm.

The scarcity of cryogenic thermometers often stems from their high cost and lengthy lead times for calibration. Establishing an in-lab temperature calibration system is necessary to quickly make use of uncalibrated sensors or self-made sensors. This paper introduces a straightforward and high-accuracy thermometer calibration system. By employing copper screws as thermal links between the sensor platform and the cryogen-free refrigerator, temperature oscillation on the sensor platform is suppressed to a few millikelvins. In addition, this paper presents a data processing model based on clustering algorithms. These algorithms sort and group data based on distance, which is similar to human visual judgment of data. This paper discusses the parameter optimization process of the clustering algorithm to interpret the automated data process. The cryogenic temperature sensors calibrated by this system exhibited high accuracy, with relative errors of less than 1% compared to standard thermometers. Moreover, automatically processing calibration data from two uncalibrated thermometers takes just over 10min, highlighting the effectiveness of this calibration system.

Anomaly detection of power grid data based on storm cloud platform sliding window

Abstract To realize the monitoring and diagnosis of multi-source data in the power industry and meet the real-time processing requirements of the power system, the storm-distributed real-time computing platform is introduced to process the data. On this platform, the streaming data processing model based on a sliding window is deployed to realize the anomaly diagnosis of data flow. By calculating the predicted value of a point and the confidence interval of the predicted value of the point, we can judge whether the actual point is an abnormal value. By testing the data processing capacity of a cluster and a single machine, the experiment shows that in the cluster environment, reasonably setting the component parallelism can improve the throughput of flow computing.

Knowledge-Augmented Deep Learning for Segmenting and Detecting Cerebral Aneurysms With CT Angiography: A Multicenter Study.

Background Deep learning (DL) could improve the labor-intensive, challenging processes of diagnosing cerebral aneurysms but requires large multicenter data sets. Purpose To construct a DL model using a multicenter data set for accurate cerebral aneurysm segmentation and detection on CT angiography (CTA) images and to compare its performance with radiology reports. Materials and Methods Consecutive head or head and neck CTA images of suspected unruptured cerebral aneurysms were gathered retrospectively from eight hospitals between February 2018 and October 2021 for model development. An external test set with reference standard digital subtraction angiography (DSA) scans was obtained retrospectively from one of the eight hospitals between February 2022 and February 2023. Radiologists (reference standard) assessed aneurysm segmentation, while model performance was evaluated using the Dice similarity coefficient (DSC). The model's aneurysm detection performance was assessed by sensitivity and comparing areas under the receiver operating characteristic curves (AUCs) between the model and radiology reports in the DSA data set with use of the DeLong test. Results Images from 6060 patients (mean age, 56 years ± 12 [SD]; 3375 [55.7%] female) were included for model development (training: 4342; validation: 1086; and internal test set: 632). Another 118 patients (mean age, 59 years ± 14; 79 [66.9%] female) were included in an external test set to evaluate performance based on DSA. The model achieved a DSC of 0.87 for aneurysm segmentation performance in the internal test set. Using DSA, the model achieved 85.7% (108 of 126 aneurysms [95% CI: 78.1, 90.1]) sensitivity in detecting aneurysms on per-vessel analysis, with no evidence of a difference versus radiology reports (AUC, 0.93 [95% CI: 0.90, 0.95] vs 0.91 [95% CI: 0.87, 0.94]; P = .67). Model processing time from reconstruction to detection was 1.76 minutes ± 0.32 per scan. Conclusion The proposed DL model could accurately segment and detect cerebral aneurysms at CTA with no evidence of a significant difference in diagnostic performance compared with radiology reports. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Payabvash in this issue.

The Impact of Seeking and Processing Environmental Information Related to Microplastics on Pro-environmental Behavior Intentions : Focusing on the Risk Information Seeking and Processing Model (RISP)

The Impact of Seeking and Processing Environmental Information Related to Microplastics on Pro-environmental Behavior Intentions : Focusing on the Risk Information Seeking and Processing Model (RISP)

Expanding the neighbourhood watch: Orthographic neighbours in isiXhosa reading and spelling

Background: Lexical properties such as orthographic neighbours have been shown to have an influence on reading and writing; however, this phenomenon is yet to be explored in the Southern Bantu languages.Objectives: We investigate the role of orthographic neighbourhood density and neighbourhood frequency in reading and spelling in Grade 3 isiXhosa home-language learners. The aim is to establish whether orthographic neighbours facilitate or hinder reading and spelling. The dual-route model of orthographic processing is used to interpret our findings.Method: Data were collected from 97 Grade 3 isiXhosa home-language learners. Lexical decision, word reading, and spelling tasks were administered. Tasks included both real and pseudowords varying in orthographic neighbourhood density and neighbourhood frequency, while controlling for word length and word frequency.Results: Findings show a significant inhibitory effect of orthographic neighbourhood frequency and word length for spelling accuracy. Longer words with higher frequency neighbours were more likely to be spelt incorrectly. No effect was recorded for the lexical decision and word reading tasks.Conclusion: Our findings suggest that spelling in isiXhosa relies partially on lexical processing strategies, which allows for the inhibitory effects of orthographic neighbours to be observed. While in reading, sublexical processing is more prominent, therefore mitigating the effect of orthographic neighbours.Contribution: Empirical evidence for the effects of orthographic neighbours, specifically in a Southern Bantu language, provides a clearer picture of the underlying cognitive-linguistic processes involved in reading and writing. This evidence may contribute to the development of targeted pedagogical practices to address spelling errors.

The potential of synthetic minority oversampling technique to enhance the precision of gender prediction: an investigation of artificial neural networks with cephalometry

BACKGROUND: When creating models utilizing artificial neural networks, the quantity of training data and the distribution of data need to be considered, particularly when making gender predictions. AIM: This study seeks to determine the potential impact of using the synthetic minority oversampling technique (SMOTE) on gender prediction using the artificial neural networks model. MATERIALS AND METHODS: The current study utilized a dataset consisting of 297 cephalometric measurements from Indonesian patients, comprising 229 samples from females and 68 samples from males. WebCeph is used to measure certain parameters, such as Sella-Nation-Point A (SNA) angle, mandibular length, mandibular angle, Sella-Glabella-Point A (SGA) angle, and diagnosis. Data processing and artificial neural networks model creation were conducted using Python. RESULTS: The gender identification accuracy of the artificial neural networks model is 87% for females and 0% for males, resulting in an overall average accuracy of 78%. When using SMOTE, the accuracy is 22%, with 0% for females and 37% for males. However, when using SMOTE and normalization, the accuracy increases to 71%, with 82% for females and 30% for males. The accuracy of normalization without SMOTE is 76%, with 86% for females and 14% for males. CONCLUSIONS: This research has proven the efficacy of SMOTE in improving the classification of male matrices. Nevertheless, this study reveals that the overall accuracy results of SMOTE are suboptimal in comparison to the absence of SMOTE and normalization. The application of data balancing strategies is necessary to achieve optimal accuracy in gender prediction when artificial neural networks, and other parameters must be applied.

Adjusting Mood in Mandarin Chinese: A Game Theory Approach to Double and Redundant Negation with Entropy

Abstract This study investigates how Mandarin Chinese speakers use double and redundant negations to adjust information entropy and convey the speaker’s mood. Integrating Gricean implicature, signaling game theory, and information entropy, we quantified uncertainty in negation sentences by calculating self-information of negation markers and overall entropy for negations versus affirmatives. Analyzing sentences from the BCC Corpus of Mandarin Chinese newspapers, we extracted double and redundant negation patterns, calculated their frequencies, and compared entropy values in payoff matrices against affirmatives. Contextual analysis of surrounding discourse interpreted pragmatic functions. Our findings reveal that universal double negations reduce sentence entropy, intensifying assertiveness. Modal forms further decrease entropy, amplifying necessity, while conditional patterns strengthen the asserted conditions. Euphemistic double negations increase entropy, weaken assertiveness, and align with Levinson’s Manner Principle. This increase in entropy quantifies pragmatic markedness, allowing tentative critiques. Redundant negations significantly reduce entropy, strengthening negative attitudes. We elucidate how interlocutors navigate cooperative discourse by quantifying uncertainty via entropy measures, making contextually-optimal linguistic choices like assertion, indirectness, and emphasis. These findings impact signaling game theory and pragmatic markedness, with applications in natural language processing, cross-cultural communication, and computational models. Future research could explore other languages and cognitive and social factors in negation use.

NetraDeep: An Integrated Deep Learning and Image Processing System for Precise Detection of Hard Exudates

Hard Exudate (HE) is a common manifestation of various eye diseases, such as diabetic retinopathy (DR), and a prominent cause of vision loss and blindness. Researchers aim to visualize and quantify these exudates using deep learning (DL) and image processing (IP) models from retinal images. However, the requirement for a large number of labelled image datasets for DL models to work on diverse and poor-quality images makes this task challenging. To address this challenge, we introduce NetraDeep, a system that integrates data-driven DL and rule-based IP techniques for exudate segmentation. Our system uses IP models to detect and extract some features and assists DL models in detecting more advanced features and vice versa. The IP models are rule-based and use predefined rules to process images, while the DL models are data-driven and learn from the input data. NetraDeep provides visual and quantitative assessments while mitigating noise and other confounding factors such as artifacts and noise. The training of DL models of this system requires only a limited number of labelled fundus images from publicly available datasets. It provides accurate pixel-wise segmentation results on the public and private image datasets collected from local eye hospitals. Through extensive evaluation, our system achieved remarkable performance, with a dice coefficient of \(0.84\) for the public dataset and a rating of \(9.78\) and \(9.43\) out of 10, as corroborated by two medical experts with experience of more than 20 and 5 years, respectively, for the private image dataset.

Language or rating scales based classifications of emotions: computational analysis of language and alexithymia

Rating scales are the dominating tool for the quantitative assessment of mental health. They are often believed to have a higher validity than language-based responses, which are the natural way of communicating mental states. Furthermore, it is unclear how difficulties articulating emotions—alexithymia—affect the accuracy of language-based communication of emotions. We investigated whether narratives describing emotional states are more accurately classified by questions-based computational analysis of language (QCLA) compared to commonly used rating scales. Additionally, we examined how this is affected by alexithymia. In Phase 1, participants (N = 348) generated narratives describing events related to depression, anxiety, satisfaction, and harmony. In Phase 2, another set of participants summarized the emotions described in the narratives of Phase 1 in five descriptive words and rating scales (PHQ-9, GAD-7, SWLS, and HILS). The words were quantified with a natural language processing model (i.e., LSA) and classified with machine learning (i.e., multinomial regression). The results showed that the language-based responses can be more accurate in classifying the emotional states compared to the rating scales. The degree of alexithymia did not influence the correctness of classification based on words or rating scales, suggesting that QCLA is not sensitive to alexithymia. However, narratives generated by people with high alexithymia were more difficult to classify than those generated by people with low alexithymia. These results suggest that the assessment of mental health may be improved by language-based responses analyzed by computational methods compared to currently used rating scales.

ChatGPT Predicts In-Hospital All-Cause Mortality for Sepsis: In-Context Learning with the Korean Sepsis Alliance Database.

Sepsis is a leading global cause of mortality, and predicting its outcomes is vital for improving patient care. This study explored the capabilities of ChatGPT, a state-of-the-art natural language processing model, in predicting in-hospital mortality for sepsis patients. This study utilized data from the Korean Sepsis Alliance (KSA) database, collected between 2019 and 2021, focusing on adult intensive care unit (ICU) patients and aiming to determine whether ChatGPT could predict all-cause mortality after ICU admission at 7 and 30 days. Structured prompts enabled ChatGPT to engage in in-context learning, with the number of patient examples varying from zero to six. The predictive capabilities of ChatGPT-3.5-turbo and ChatGPT-4 were then compared against a gradient boosting model (GBM) using various performance metrics. From the KSA database, 4,786 patients formed the 7-day mortality prediction dataset, of whom 718 died, and 4,025 patients formed the 30-day dataset, with 1,368 deaths. Age and clinical markers (e.g., Sequential Organ Failure Assessment score and lactic acid levels) showed significant differences between survivors and non-survivors in both datasets. For 7-day mortality predictions, the area under the receiver operating characteristic curve (AUROC) was 0.70-0.83 for GPT-4, 0.51-0.70 for GPT-3.5, and 0.79 for GBM. The AUROC for 30-day mortality was 0.51-0.59 for GPT-4, 0.47-0.57 for GPT-3.5, and 0.76 for GBM. Zero-shot predictions using GPT-4 for mortality from ICU admission to day 30 showed AUROCs from the mid-0.60s to 0.75 for GPT-4 and mainly from 0.47 to 0.63 for GPT-3.5. GPT-4 demonstrated potential in predicting short-term in-hospital mortality, although its performance varied across different evaluation metrics.

A wavelet packet deep learning model for Energy-Based structural collapse assessment under Earthquake-Fire Scenarios: Framework and hybrid simulation

An energy-based framework is proposed for the dynamic stability assessment of structures subjected to earthquake-fire scenarios and verified through earthquake-fire hybrid simulation. In this framework, the wavelet packet Long Short-Term Memory (LSTM) model is used to separate the noise and residue caused by earthquake and fire within the structural response signals, guided by wavelet packet energy and power spectral density of signals. Moreover, a comparative analysis is performed with 4 previous signal processing models (empirical mode decomposition, variational mode decomposition, wavelet packet transform, and LSTM). Additionally, Latin hypercube sampling is employed to account for uncertainties in structural characteristics and hazards for dataset establishment and fragility analysis. The experimental findings suggest a decline in the structural dominant frequency with the emergence of high-frequency noise and residuals in structural responses due to the multi-hazard effect. The wavelet packet transform eliminates the high-frequency noise in the signal and avoids the oscillation of the LSTM prediction results. Post-earthquake fire increases the structural collapse possibility even under a moderate earthquake excitation. The proposed framework proves to be rational and has the potential to be further applied to other multi-hazard scenarios.

Financial Revolution through Agent-based Artificial Simulation Computational Models for Predicting Market Behavior

The fundamental theory of the Efficient market hypothesis (EMH), which states that market participants are rational, has received a lot of criticism. The complexity of behavior in the capital market is still a black box, especially when psychological biases influence aggressively on decision-making amid uncertainty. Experimental research on finance and capital markets in the form of AI using machine learning seeks to predict the results of more complex interactions. This multidisciplinary approach offers efforts to explain social phenomena from the micro level to macro descriptions which are built artificially through the computational world. The processing modeling approach is preferred because it includes the complexes that emerge from the behavior and interactions of individuals in the real world. Agent Based Model (ABM) is an AI approach in the form of computational simulation that performs a bottom-up approach by combining irrational–rational agent interactions through networks in microenvironments. Using the ABM approach through Netlogo computing, this study proves that AI can be used to analyze investor behavior in the capital market. Keywords: Agent Based Model, artificial intelligence, investor behavior

OPTIMIZATION OF WASTE MANAGEMENT IN SUMBAWA DISTRICT USING A DYNAMIC SYSTEM MODEL

The main problem that occurs in society today apart from basic needs is the problem of waste. This waste problem certainly needs a solution in good, effective, sustainable and long-term management. The waste issue not only occur in big cities but also in developing areas. Sumbawa Regency as one of the regions that continues to develop and grow requires optimum and long-term planning of solid waste management and systems. In 2021, the waste handled in Sumbawa Regency will only be around 17.5% of the total waste. This shows that the waste management system in Sumbawa Regency is not optimal. Optimal waste management can be achieved by taking an academic approach through model development. One such model is the dynamic system model. This can describe the causal relationship between variables that affect solid waste problems and to provide an overview of the waste management system through simulations based on the factors that influence the processing model. Therefore, it is necessary to develop an optimization model for waste management specifically in Sumbawa Regency using a dynamic systems approach. The purpose of this study is to describe waste management in the form of an optimization model for waste management in Sumbawa Regency with a dynamic system. Optimization results show that the optimistic model yields a total of 127.01 tons/day, the moderate model yields a total of 98.23 tons/day and the pessimistic model produces 62.5 tons/day

Evaluating automatic annotation of lexicon-based models for stance detection of M-pox tweets from May 1st to Sep 5th, 2022.

Manually labeling data for supervised learning is time and energy consuming; therefore, lexicon-based models such as VADER and TextBlob are used to automatically label data. However, it is argued that automated labels do not have the accuracy required for training an efficient model. Although automated labeling is frequently used for stance detection, automated stance labels have not been properly evaluated, in the previous works. In this work, to assess the accuracy of VADER and TextBlob automated labels for stance analysis, we first manually label a Twitter, now X, dataset related to M-pox stance detection. We then fine-tune different transformer-based models on the hand-labeled M-pox dataset, and compare their accuracy before and after fine-tuning, with the accuracy of automated labeled data. Our results indicated that the fine-tuned models surpassed the accuracy of VADER and TextBlob automated labels by up to 38% and 72.5%, respectively. Topic modeling further shows that fine-tuning diminished the scope of misclassified tweets to specific sub-topics. We conclude that fine-tuning transformer models on hand-labeled data for stance detection, elevates the accuracy to a superior level that is significantly higher than automated stance detection labels. This study verifies that automated stance detection labels are not reliable for sensitive use-cases such as health-related purposes. Manually labeled data is more convenient for developing Natural Language Processing (NLP) models that study and analyze mass opinions and conversations on social media platforms, during crises such as pandemics and epidemics.