Subjective Bias Research Articles

A hybrid fault diagnosis scheme for milling tools using MWN-CBAM-PatchTST network with acoustic emission signals

ABSTRACT Milling tools are critical to machining and manufacturing processes. Accurate diagnosis and identification of faults occurring in milling tools during their operation are of utmost importance for maintaining the reliability and availability of these tools, and to minimise machine downtime and overall machining costs. This paper presents a milling tools fault diagnosis network model based on acoustic emission signals. The model integrates a multilayer wavelet CNN (MWN) consisting of a discrete wavelet transform (DWT) and a convolutional neural network (CNN), a convolutional block attention module (CBAM), and a PatchTST module. MWN uses wavelet transformation to withdraw multi-scale features from signals, thus improving sensitivity to small variations in acoustic emission. CBAM improves feature representation by focusing on critical feature channels and regions, while PatchTST uses a self-attention mechanism to optimise processing of long-range feature dependencies. This synergy of mechanisms results in superior performance, outperforming traditional diagnostic methods. Bayesian optimisation is used to select model hyperparameters, eliminating the subjective bias associated with manual range setting. Validation experiments using the milling dataset, including ablation studies and comparative tests, demonstrated that the model achieves an identification accuracy of over 98%, validating its generalisation capability and effectiveness in diagnosing tool faults using acoustic emission signals.

Advancing precision agriculture with deep learning enhanced SIS-YOLOv8 for Solanaceae crop monitoring

IntroductionPotatoes and tomatoes are important Solanaceae crops that require effective disease monitoring for optimal agricultural production. Traditional disease monitoring methods rely on manual visual inspection, which is inefficient and prone to subjective bias. The application of deep learning in image recognition has led to object detection models such as YOLO (You Only Look Once), which have shown high efficiency in disease identification. However, complex climatic conditions in real agricultural environments challenge model robustness, and current mainstream models struggle with accurate recognition of the same diseases across different plant species.MethodsThis paper proposes the SIS-YOLOv8 model, which enhances adaptability to complex agricultural climates by improving the YOLOv8 network structure. The research introduces three key modules: 1) a Fusion-Inception Conv module to improve feature extraction against complex backgrounds like rain and haze; 2) a C2f-SIS module incorporating Style Randomization to enhance generalization ability for different crop diseases and extract more detailed disease features; and 3) an SPPF-IS module to boost model robustness through feature fusion. To reduce the model’s parameter size, this study employs the Dep Graph pruning method, significantly decreasing parameter volume by 19.9% and computational load while maintaining accuracy.ResultsExperimental results show that the SIS-YOLOv8 model outperforms the original YOLOv8n model in disease detection tasks for potatoes and tomatoes, with improvements of 8.2% in accuracy, 4% in recall rate, 5.9% in mAP50, and 6.3% in mAP50-95.DiscussionThrough these network structure optimizations, the SIS-YOLOv8 model demonstrates enhanced adaptability to complex agricultural environments, offering an effective solution for automatic crop disease detection. By improving model efficiency and robustness, our approach not only advances agricultural disease monitoring but also contributes to the broader adoption of AI-driven solutions for sustainable crop management in diverse climates.

The use of deep learning in intelligent athlete motion recognition: Integrating biological mechanisms

This work explores the effective application of deep learning for recognizing athletes’ movements, aiming to enhance precision in competitive sports. Traditional motion analysis methods primarily rely on manual observation, which can introduce subjective bias and limit accuracy. To address these limitations, we propose an automated method based on deep learning for recognizing and classifying athletes’ technical movements while evaluating their performance. A hybrid model, combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, is utilized to extract key frames from video data. The CNN is responsible for feature extraction, capturing the intricate details of movement, while the LSTM captures the temporal sequence characteristics, providing context to the actions. To further strengthen our approach, we delve into the biological mechanisms underlying athletic movements. Understanding the biomechanics of motion—such as joint angles, muscle activation patterns, and energy expenditure—can enhance the accuracy of deep learning models. By integrating these biological insights into our model, we improve the recognition process, allowing for a more nuanced understanding of how movements impact performance. Through experiments, we demonstrate that the model achieves high accuracy across multiple benchmark datasets (UCF-101, HMDB-51, Kinetics-400, and Sports-1M), with a particularly high accuracy of 93.5% on the UCF-101 dataset. These results indicate that the proposed method is both accurate and reliable, making it suitable for athlete training and competition analysis. The findings of this research have significant implications for sports science, training evaluation, and injury prevention. By providing coaches and athletes with precise feedback based on deep learning analysis, we can facilitate targeted training interventions that enhance performance while reducing injury risks. This work aims to offer a powerful tool for athletes, coaches, and researchers, contributing to the advancement of competitive sports through a deeper understanding of movement dynamics and their biological underpinnings.

A novel method for objective quantification of apathy based on gaze and physiological reactivity to stimuli presented in a virtual reality environment

AbstractINTRODUCTIONWe developed a tool for objective quantification of apathy.METHODSParticipants (n = 97; 67 with cognitive impairment, 30 cognitively normal; mean age = 74.3 ± 6.2 years, 56.7% females) were exposed to emotional and cognitive stimuli in a virtual reality environment. Gaze metrics (time to first fixation [TTFF] and total fixation duration [TFD]) and autonomic nervous system (ANS) reactivity were measured. Apathy and depression were clinically assessed using the Lille Apathy Rating Scale short version and the Geriatric Depression Scale 15‐item version, respectively. Cutoffs of ≥ –7 and ≥ 5 were used to define apathy and depression, respectively.RESULTSThe sample comprised 14 participants with apathy only, 9 with depression only, 10 with both, 63 with neither, and 1 with missing data. For all emotional stimuli, participants with apathy only showed longer TTFF (P = 0.039, effect sizes [ES] = 0.798), and shorter TFD (P = 0.023, ES = 0.578) compared to those without apathy or depression. ANS reactivity was not associated with apathy.DISCUSSIONApathy is associated with decreased gaze engagement at emotional stimuli.Highlights Apathy measurement via questionnaires is limited by subjectivity biases. Apathy measurement via questionnaires is limited by simplistic scoring. We present a novel method for objective measurement of apathy. Gaze characteristics reflect the emotional and cognitive components of apathy.

Plasmon Dynamics in Nanoclusters: Dephasing Revealed by Excited States Evaluation.

The photocatalytic efficiency of materials such as graphene and noble metal nanoclusters depends on their plasmon lifetimes. Plasmon dephasing and decay in these materials is thought to occur on ultrafast time scales, ranging from a few femtoseconds to hundreds of femtoseconds and longer. Here we focus on understanding the dephasing and decay pathways of excited states in small lithium and silver clusters and in plasmonic states of the π-conjugated molecule anthracene, providing insights that are crucial for interpreting optical properties and photophysics. To do this, we study the time dependence of the electronic density matrix of these molecules using a new approach that expresses the density matrix in terms of TDDFT eigenstates (ESs) of the TDDFT Hamiltonian. This approach, which involves combining linear response time-dependent density functional theory (LR-TDDFT) and real-time time-dependent density functional theory (RT-TDDFT), leads to an analysis of the electron dynamics in terms of ESs, rather than individual molecular orbital (MO) transitions as has typically been done. This circumvents the complexities and subjective biases that traditional MO-based analysis provides. We find in an analysis of the induced dipole moment in these molecules that what had previously been considered to be energy relaxation is actually dephasing associated with the eigenstates that are stationary after the excitation pulse is turned off. We conclude that the ES-basis analysis has significant potential to advance understanding of the electron dynamics of plasmonic nanomaterials, aiding their optimization for photocatalytic and technological applications.

Social interaction models for trust systems design

AbstractFueled by the social dynamics inherent in traditional networking, the Internet has come to shape a new paradigm of social interaction and collaboration, formed by the active engagement of its users. The emergence of social networks has not only expanded the avenues for blending the social essence of trust, but also harnessed the technical capabilities of the Internet. While many computational trust models employ models from social theories of interaction to accurately identify trustworthy individuals, there is a notable gap in addressing how these models are perceived and acted upon by the users, who are the core of trust systems. Recognizing and understanding the relationship between trust models and user perceptions is key to simplifying their complexity and enhancing model explainability and system performance.This paper delves into the nuanced roles of structural balance, subjectivity, and user bias within trust systems, investigating their implications on the design of computational trust models. Through qualitative and quantitative analyses, we pinpoint instances where the nature of social interactions emerges as a significant concern that needs preemptive consideration during the design phase of trust systems. Building upon the findings, we suggest practical solutions to enhance the effectiveness of trust models through alignment between the contextual system traits and the design assumptions.

Metamodeling of turbofan engine compressor characteristics using SVM-based improved Kriging method

Purpose The purpose of this paper is to overcome the inherent lack of precision in commonly used interpolation procedures when solving the mathematical model of turbofan engines, as well as to address the issue that the theoretical variogram model in traditional Kriging models is prone to subjective selection bias, which makes it impossible to accurately capture the inherent fluctuation patterns in compressor data. Design/methodology/approach To mitigate this challenge, based on the spatial distribution characteristics of the compressor characteristic data of a certain type of turbofan engine, the input and output dimensions of the model are defined. By determining the stable operating region from the original component data, the authors use the proposed Kriging method improved with a support vector machine model to reconstruct the characteristics at unknown speeds within this region. The effectiveness of the proposed method is evaluated using the established assessment metrics. Findings Experimental results demonstrate that the proposed method exhibits significant advantages over the conventional Kriging approach. Specifically, it leads to a substantial reduction in root mean square error and mean absolute error by 0.0153/0.0118 (low speed), 0.1306/0.0362 (medium speed) and −0.0066/0.2366 (high speed). Originality/value This refined approach not only offers notable engineering applicability but also contributes significantly to the enhancement of aerospace engine model solutions’ precision.

Leveraging Machine Learning for Talent Acquisition: Predicting High-Performance Candidates in Human Resource Management

This study explores the application of machine learning (ML) in human resource (HR) management to enhance the recruitment process by predicting high-performing candidates. The research addresses gaps in traditional recruitment methods, which are often time-consuming and susceptible to subjective bias. By employing a Random Forest algorithm, this study utilizes a dataset of 10,000 records, encompassing attributes such as education, work experience, psychometric assessments, and interview evaluations. Data were divided into 70% training and 30% testing sets to ensure robust model evaluation. The findings demonstrate that the Random Forest model achieved a prediction accuracy of 87%, outperforming traditional methods and other ML models like Logistic Regression. The model's ability to identify key attributes contributing to candidate performance underscores its potential for data-driven decision-making in HR management. However, challenges such as data bias, algorithmic transparency, and resistance to technological change were identified as barriers to implementation. This research contributes to the theoretical and practical understanding of ML in HR by offering a predictive model that balances accuracy with interpretability. Practical implications include strategies for integrating ML into existing HR systems, emphasizing the importance of explainable AI to foster trust among practitioners. The study concludes that ML-based recruitment can significantly improve efficiency, objectivity, and the quality of hiring decisions, paving the way for more innovative and strategic HR practices.

The Comparison and Selection of a Warm Mix Agent for Asphalt Based on Fuzzy Hierarchical Analysis

Warm mix asphalt pavements can effectively reduce construction energy consumption and control environmental pollution compared to conventional asphalt pavements. However, there is no clear test standard for the evaluation of warm mix agents. The selection of a warm mix agent is easily influenced by subjective factors. Moreover, it is difficult to consider quantitative and qualitative factors such as technical indexes and the economy. A fuzzy hierarchical comprehensive evaluation model of warm mix agents is proposed based on fuzzy hierarchical analysis. Four types of indexes, namely, material properties, technical indexes, economy, and construction difficulty, were selected to construct the hierarchical structure of the target, guideline, and indicator layers. The influence weights of all factors were determined by the judgment matrix, and an affiliation judgment matrix was established at the same time. Finally, the comprehensive score of the warm mix agent modification effect was calculated. Three warm mix agents were selected for market research and indoor testing. And the evaluation model was applied to process the data. The calculation results show that the technical indicators have the greatest influence, with a weight proportion of 0.493; WM-2 had the best comprehensive evaluation, with a comprehensive score of 0.96, and its optimal mixing amount was 0.3%. The calculation results further verified the effectiveness of the fuzzy hierarchical comprehensive evaluation model of warm mix agents. Compared with the existing methods, the fuzzy-hierarchical-integrated evaluation model of warm mix agents has a more comprehensive evaluation scope, lower subjective factor bias, and more flexible application scenarios. It provides a more rigorous and convenient idea for the development and evaluation of different warm mix agents. In conclusion, the model has greater application value and popularization value.

Effect of Level and Frequency of Forward Masker on Auditory Brainstem Response.

Forward masking (FM) is characterized by the perception of a signal being reduced or wholly masked due to a preceding sound (masker) of the same or different frequencies that offers a challenge for the auditory system to resolve. Considering that the off-frequency masker is expected to undergo linear processing compared to the on-frequency masker at the signal place, it reflects the peripheral auditory systems' compressive response. Thus, the present study focused on employing FM electrophysiological analogous such as auditory brainstem responses (ABR) to the behavioral masking experiments to objectively measure the frequency and level of processing in the auditory system, from the periphery to the brainstem level. The study was an observational research on 21 female volunteers. ABR was obtained using a tone-on-tone FM paradigm for 1000- and 4000-Hz probe stimuli. An experiment used two forward maskers, on-frequency and off-frequency, with varying levels from 50 to 70 dB SPL. A progressive shift for Vth peak latency and reduction in response amplitude was observed in proportion to the increase of masker level for both the probe stimuli and the masking experiments. However, ABR responses in neither masking condition were observed to differ between 60 and 70 dB SPL. FM ABR experiments are an assessment tool for estimating frequency and level processing in the auditory system, providing good efficiency, reliability, and less subject bias compared to behavioral measures.

Validating automated resonance evaluation with synthetic data

The integrity and precision of nuclear data are crucial for a broad spectrum of applications, from national security and nuclear reactor design to medical diagnostics, where the associated uncertainties can significantly impact outcomes. A substantial portion of uncertainty in nuclear data originates from the subjective biases in the evaluation process, a crucial phase in the nuclear data production pipeline. Recent advancements indicate that automation of certain routines can mitigate these biases, thereby standardizing the evaluation process and enhancing reproducibility. This research aims to provide a methodology, framework, and metrics for the validation of automated nuclear data evaluation software leveraging high-quality synthetic data that closely mimic real experimental observables. An introduced error metric provides a scale and intuitive measure of the evaluation quality by quantifying the estimate’s accuracy and performance across the specified energy range. Synthetic data provides access to experimental observables and underlying resonance parameters, enabling comparison of different evaluations. The methodology is demonstrated using Ta-181 isotope data in the resolved resonance region. The Automated Resonance Identification Subroutine (ARIS), which operates without prior resonance information, was used to test and showcase the framework’s capabilities utilizing the proposed error metrics. The results demonstrate the effectiveness of the proposed approach and framework for optimizing software parameters and testing hypotheses through “what-if” controlled experiments, such as modifying assumptions about experimental conditions or average resonance parameters.

Compositional effects on the etching of fossil confined fission tracks in apatite

Abstract Fission track analysis is a thermochronologic method for dating rocks and reconstructing their low-temperature thermal histories. We investigate the influence of the apatite composition on the etching of fossil confined fission tracks and its consequences for the fission track method. We conducted step-etch experiments with 5.5 M HNO3 at 21 °C on samples with etch pit diameters (Dpar) spanning most of the range for natural apatites (Panasqueira: 1.60 µm; Slyudyanka: 2.44 µm; Brazil: 3.92 µm; and Bamble: 4.60 µm) to determine their apatite etch rates vR (the rate at which each lattice plane is displaced parallel to itself) as a function of crystallographic orientation (ϕ′). Our measurements revealed significant differences between the four samples. We fitted three-parameter functions, vR = a(Dpar)ϕ′ eb(Dpar)ϕ′ + c, describing vR as a function of the angle to the apatite c-axis for our hexagonal samples (excluding Bamble) and Durango apatite. Both parameters a and b exhibit a linear correlation with Dpar, whereas the constant c is small (~0.1 µm/min) and its between-sample variation is negligible at the resolution of our measurements. Bamble exhibits a different, bimodal relationship between vR and ϕ′, which we fitted with a sum of two sine functions. In all cases, including Bamble, there is a striking correlation between the angular frequencies of horizontal confined tracks and the magnitude of the apatite etch rate vR perpendicular to the track axes. This result shows that the sample of confined tracks selected for measurement and modeling is to a much greater degree determined by the etching properties of the apatite sample than by geometric or subjective biases. The track etch rate vT is constant along most of the track length but varies from track to track. The mean vT correlates with Dpar, so that tracks etch to their full lengths in a shorter time in faster etching apatites. The mean rate of length increase between etch steps, vL, also correlates with Dpar. The length increments of individual tracks are however irregular. This points to an intermittent structure at the ends of the tracks.

SLA-MLP: Enhancing Sleep Stage Analysis from EEG Signals Using Multilayer Perceptron Networks.

Background/Objectives: Sleep stage analysis is considered to be the key factor for understanding and diagnosing various sleep disorders, as it provides insights into sleep quality and overall health. Methods: Traditional methods of sleep stage classification, such as manual scoring and basic machine learning approaches, often suffer from limitations including subjective biases, limited scalability, and inadequate accuracy. Existing deep learning models have improved the accuracy of sleep stage classification but still face challenges such as overfitting, computational inefficiencies, and difficulties in handling imbalanced datasets. To address these challenges, we propose the Sleep Stage Analysis with Multilayer Perceptron (SLA-MLP) model. Results: SLA-MLP leverages advanced deep learning techniques to enhance the classification of sleep stages from EEG signals. The key steps of this approach include data collection, where diverse and high-quality EEG data are gathered; preprocessing, which involves signal cropping, spectrogram conversion, and normalization to prepare the data for analysis; data balancing, where class weights are adjusted to address any imbalances in the dataset; feature extraction, utilizing Temporal Convolutional Networks (TCNs) to extract meaningful features from the EEG signals; and final classification, applying a Multilayer Perceptron (MLP) to accurately predict sleep stages. Conclusions: SLA-MLP demonstrates superior performance compared to traditional methods by effectively addressing the limitations of existing models. Its robust preprocessing techniques, advanced feature extraction, and adaptive data balancing strategies collectively contribute to obtaining more accurate results, having an accuracy of 97.23% for the S-DSI, 96.23 for the S-DSII and 97.23% for the S-DSIII dataset. This model offers a significant advancement in the field, providing a more precise tool for sleep research and clinical applications.

SaltFormer: A hybrid CNN-Transformer network for automatic salt dome detection

Salt dome interpretation of seismic data is a crucial task in the exploration and development of oil and gas. Conventional techniques, such as multi-attribute analysis, are laborious, time-consuming, and susceptible to subjective biases in their results. To achieve a more automated and precise identification of salt dome, we developed a hybrid network for salt dome detection. In order to optimally exploit both local and global features, a hierarchical Vision Transformer is employed as an encoder for feature extraction. Concurrently, the concurrent spatial and channel squeeze & excitation attention module is utilized to improve detection accuracy in the decoder. Furthermore, we leveraged the complementarity of information between multiple tasks to enhance the model’s generalization performance. Using the competition data from the Kaggle platform provided by TGS-NOPEC Geophysics Company, automatic segmentation of salt domes was completed with a detection accuracy of 85.20%. A series of experiments were conducted using state-of-the-art models and the SaltFormer model, which was found to have higher detection accuracy compared to other networks. Finally, the test conducted with seismic field data from the Netherlands offshore F3 block in the North Sea demonstrate that the novel method is highly effective in detecting salt domes in seismic data.

Pronoun interpretation is more subject-biased than expected by the Bayesian Model

ABSTRACT A Bayesian Model proposed by Kehler et al. ([2008]. Coherence and coreference revisited. Journal of Semantics, 25(1), 1–44. https://doi.org/10.1093/jos/ffm018) suggests that pronoun production and interpretation are driven by a different set of factors following the Bayes’ rule. Evidence suggests that the Bayesian Model makes better predictions on pronoun interpretation compared to other models that assume that pronoun production and interpretation are influenced by the same set of factors. Yet, it remains unclear precisely to what extent the Bayesian Model can capture this relationship. The current study examines the validity of the Bayesian Model by comparing its performance across three different contexts using a variety of evaluation methods. Our results demonstrate that the Bayesian Model’s performance varied across contexts and consistently underestimated the subject bias in interpretation. We suggest that the underestimation is likely because the subject bias in pronoun production is not sufficient to account for the subject bias in actual interpretation, contra to the assumption of the Bayesian Model. We discuss potential sources of the additional subject bias in interpretation.

The use of 4D data-independent acquisition-based proteomic analysis and machine learning to reveal potential biomarkers for stress levels.

Research suggests that individuals who experience prolonged exposure to stress may be at higher risk for developing psychological stress disorders. Currently, psychological stress is primarily evaluated by professional physicians using rating scales, which may be prone to subjective biases and limitations of the scales. Therefore, it is imperative to explore more objective, accurate, and efficient biomarkers for evaluating the level of psychological stress in an individual. In this study, we utilized 4D data-independent acquisition (4D-DIA) proteomics for quantitative protein analysis, and then employed support vector machine (SVM) combined with SHAP interpretation algorithm to identify potential biomarkers for psychological stress levels. Biomarkers validation was subsequently achieved through machine learning classification and a substantial amount of a priori knowledge derived from the knowledge graph. We performed cross-validation of the biomarkers using two batches of data, and the results showed that the combination of Glyceraldehyde-3-phosphate dehydrogenase and Fibronectin yielded an average area under the curve (AUC) of 92%, an average accuracy of 86%, an average F1 score of 79%, and an average sensitivity of 83%. Therefore, this combination may represent a potential approach for detecting stress levels to prevent psychological stress disorders.

Theories of change: navigating diverse expert perceptions and preferences for global food system transformation

IntroductionEfforts are underway to transform food systems in light of their contributions to global challenges like climate change. However, food systems are highly complex, involve noteworthy place-based challenges, and there is often debate and disagreement among experts over appropriate technologies or interventions to prioritize. Tracking progress, and understanding these differences, is thus a critical need.MethodsWe surveyed food systems experts in eight countries about their preferences for 20 different food system transformation strategies and their sentiment regarding whether current initiatives are sufficient to meet 2030 goals for climate and biodiversity.ResultsExpert sentiment is overwhelmingly negative, and experts are concerned about multiple “transformation gaps,” including gaps in ambition, strategy, and implementation. Expert rankings for 20 strategies vary notably among countries and in ways that do not match those same experts’ rankings for the strength of the science behind each lever. Factor analysis reveals four distinct theories of change informing experts’ subjective biases: transformation via technical optimization, via smallholder support, via nature-positive solutions, and via supply chain enabling conditions.DiscussionThese findings provide insights for navigating the complexities of food system transformation and illustrate the influence on our strategies of preconceptions and biases in how we have come to understand the nature of the challenge.

Measuring complexity in mega construction projects: fuzzy comprehensive evaluation and grey relational analysis

PurposeMega construction projects (MCPs), characterized by their vast scale, numerous stakeholders and complex management, often face significant uncertainties and challenges. While existing research has explored the complexity of MCPs, it predominantly focuses on qualitative analysis and lacks systematic quantitative measurement methods. Therefore, this study aims to construct a complexity measurement model for MCPs using fuzzy comprehensive evaluation and grey relational analysis.Design/methodology/approachThis study first constructs a complexity measurement framework through a systematic literature review, covering six dimensions of technical complexity, organizational complexity, goal complexity, environmental complexity, cultural complexity and information complexity and comprising 30 influencing factors. Secondly, a fuzzy evaluation matrix for complexity is constructed using a generalized bell-shaped membership function to effectively handle the fuzziness and uncertainty in the assessment. Subsequently, grey relational analysis is used to calculate the relational degree of each complexity factor, identifying their weights in the overall complexity. Finally, the weighted comprehensive evaluation results of project complexity are derived by combining the fuzzy evaluation results with the grey relational degrees.FindingsTo validate the model’s effectiveness, the 2020 Xi’an Silk Road International Conference Center construction project is used as a case study. The results indicate that the overall complexity level of the project is moderate, with goal complexity being the highest, followed by organizational complexity, environmental complexity, technical complexity, cultural complexity and informational complexity. The empirical analysis demonstrates that the model can accurately reflect the variations across different dimensions of MCP complexity and can be effectively applied in real-world projects.Originality/valueThis study systematically integrates research on MCPs complexity, establishing a multidimensional complexity measurement framework that addresses the limitations of previous studies focusing on partial dimensions. Moreover, the proposed quantitative measurement model combines fuzzy comprehensive evaluation and grey relational analysis, enhancing the accuracy and objectivity of complexity measurement while minimizing subjective bias. Lastly, the model has broad applicability and can be used in MCPs across different countries and regions, providing a scientific and effective basis for identifying and managing MCP complexity.

Precision harvesting: comparative analysis of machine learning and generative AI-based classifiers for guava fruit maturity assessment using thermal imaging

ABSTRACT Guava is a highly nutritious fruit with abundance of health benefits and is medically recommended for daily consumption. Therefore, quality assessment of guava is of significance. Manual quality grading of fruits is being performed for a long time which suffers from subjective biases. Consequently, development of a computer vision-based automated system for quality grading of fruits is essential. In this paper, maturity assessment of guava is performed using various machine learning classifiers, including neural network, fuzzy logic system, and generative AI models. Thermal images of guava are used for maturity assessment. Performance of these models is evaluated by determining the confusion matrix and classification report, considering both class-wise and overall classification. After careful observation, FLS-based ANFIS is highly recommended for grading the guava on the basis of its maturity level. Furthermore, thermal imaging is also very significant for the development of a holistic computer vision-based fruit quality assessment model.

Interpretation knowledge extraction for genetic testing via question-answer model

BackgroundSequencing-based genetic testing is widely used in biomedical research, including pathogenic microorganism detection with metagenomic next-generation sequencing (mNGS). The application of sequencing results to clinical diagnosis and treatment relies on various interpretation knowledge bases. Currently, the existing knowledge bases are primarily built through manual knowledge extraction. This method requires professionals to read extensive literature and extract relevant knowledge from it, which is time-consuming and costly. Furthermore, manual extraction unavoidably introduces subjective biases. In this study, we aimed to automatically extract knowledge for interpreting mNGS results.MethodWe propose a novel approach to automatically extract pathogenic microorganism knowledge based on the question-answer (QA) model. First, we construct a MicrobeDB dataset since there is no available pathogenic microorganism QA dataset for training the model. The created dataset contains 3,161 samples from 618 published papers covering 224 pathogenic microorganisms. Then, we fine-tune the selected baseline model based on MicrobeDB. Finally, we utilize ChatGPT to enhance the diversity of training data, and employ data expansion to increase training data volume.ResultsOur method achieves an Exact Match (EM) and F1 score of 88.39% and 93.18%, respectively, on the MicrobeDB test set. We also conduct ablation studies on the proposed data augmentation method. In addition, we perform comparative experiments with the ChatPDF tool based on the ChatGPT API to demonstrate the effectiveness of the proposed method.ConclusionsOur method is effective and valuable for extracting pathogenic microorganism knowledge.