Classification Model Research Articles

Transitioning from Simulation to Reality: Applying Chatter Detection Models to Real-World Machining Data

Chatter, a self-excited vibration phenomenon, is a critical challenge in high-speed machining operations, affecting tool life, product surface quality, and overall process efficiency. While machine learning models trained on simulated data have shown promise in detecting chatter, their real-world applicability remains uncertain due to discrepancies between simulated and actual machining environments. The primary goal of this study is to bridge the gap between simulation-based machine learning models and real-world applications by developing and validating a Random Forest-based chatter detection system. This research focuses on improving manufacturing efficiency through reliable chatter detection by integrating Operational Modal Analysis (OMA), Receptance Coupling Substructure Analysis (RCSA), and Transfer Learning (TL). The study applies a Random Forest classification model trained on over 140,000 simulated machining datasets, incorporating techniques like Operational Modal Analysis (OMA), Receptance Coupling Substructure Analysis (RCSA), and Transfer Learning (TL) to adapt the model for real-world operational data. The model is validated against 1600 real-world machining datasets, achieving an accuracy of 86.1%, with strong precision and recall scores. The results demonstrate the model’s robustness and potential for practical implementation in industrial settings, highlighting challenges such as sensor noise and variability in machining conditions. This work advances the use of predictive analytics in machining processes, offering a data-driven solution to improve manufacturing efficiency through more reliable chatter detection.

Insight into the main determinants of the struggle against overweight and obesity in Chile: Use of random forest techniques and econometric models.

Overweight and obesity are considered the greatest public health problem in this emerging country, which worldwide has the second-highest percentage of overweight people among its population. The objective of this work was to analyse to what extent factors traditionally used in the study of overweight and obesity (such as sociodemographic and behavioural) and new variables proposed in the literature (such as stress, financial stress and emotional support) explain this disease in the adult population of Chile. Data were obtained from the III National Health Survey (ENS) administered by the Ministry of Health of Chile in 2017. The ENS collected a large amount of data with extensive geographic coverage. The survey comprised 4 questionnaires with a total of 576 questions, which were applied to a representative sample of the population in Chile. A double complementary methodological approach was adopted. A random forest (RF) classification model was used, and based on the results obtained, an econometric model of the censored dependent variable, specifically the Heckman sample selection model, was specified and estimated. The RF results allowed, for each of the factors considered in the research, the selection of variables with the greatest power to classify the individuals in the sample on the basis of nutritional state (normal weight, overweight or obese). Subsequently, the estimation of the parameters of the Heckman model made it possible to quantify the variables that most affected overweight and obesity. Most of the variables that make up the factors were found to be significant. Interestingly, psychosocial variables effectively influence overweight and obesity. In addition, the results for reviewing nutritional information and reviewing food warnings allow us to reflect on the impact that recent food policies have had on the Chilean population. The combination of RF and an econometric model allowed us to capitalize on the strength of both models to better explain the complex phenomenon of overweight and obesity. This approach allowed us to more accurately confirm the impact of traditional factors on overweight and obesity but to show also that other psychosocial factors are relevant and should be consider in future studies.

Enhancing the Recognition of Collinear Building Patterns by Shape Cognition Based on Graph Neural Networks

ABSTRACT Building patterns are important components of urban structures and functions, and their accurate recognition is the foundation of urban spatial analysis, cartographic generalization, and other tasks. Current building pattern recognition methods are often based on a shape index that can only characterize shape features from one aspect, resulting in significant errors. In this study, a building pattern recognition method based on a graph neural network is proposed to enhance shape cognition and focus on recognizing collinear patterns. First, a building shape classification model that integrates global shape and graph node structure features was constructed to quantitatively study shape cognition. Subsequently, a collinear pattern recognition (CPR) model was established based on a dual building graph. The shape cognition results were integrated into the model to enhance its recognition ability. The results show that the shape classification model can be used to effectively distinguish different shape categories and support building pattern recognition tasks. Based on the CPR model, false recognitions can be avoided, and recognition results similar to those of visual cognition can be obtained. Compared with the comparative methods, both models have significant advantages in terms of statistical results and implementation.

Near-infrared spectroscopy used in forensic wood research

The article is devoted to topical issues of increasing the efficiency of forensic wood research through the use of innovative technologies. It evaluates the possibilities of using spectroscopy in the near-infrared region to establish (confirm) the species of wood, and examines the main chemometric algorithms for creating classification models in the case of multi-object and binary classification. It is shown that an important condition for the correct interpretation of spectral data is the presence of a constantly updated database of standard (reference) NIR spectra.

MediScan: A Framework of U-Health and Prognostic AI Assessment on Medical Imaging

With technological advancements, remarkable progress has been made with the convergence of health sciences and Artificial Intelligence (AI). Modern health systems are proposed to ease patient diagnostics. However, the challenge is to provide AI-based precautions to patients and doctors for more accurate risk assessment. The proposed healthcare system aims to integrate patients, doctors, laboratories, pharmacies, and administrative personnel use cases and their primary functions onto a single platform. The proposed framework can also process microscopic images, CT scans, X-rays, and MRI to classify malignancy and give doctors a set of AI precautions for patient risk assessment. The proposed framework incorporates various DCNN models for identifying different forms of tumors and fractures in the human body i.e., brain, bones, lungs, kidneys, and skin, and generating precautions with the help of the Fined-Tuned Large Language Model (LLM) i.e., Generative Pretrained Transformer 4 (GPT-4). With enough training data, DCNN can learn highly representative, data-driven, hierarchical image features. The GPT-4 model is selected for generating precautions due to its explanation, reasoning, memory, and accuracy on prior medical assessments and research studies. Classification models are evaluated by classification report (i.e., Recall, Precision, F1 Score, Support, Accuracy, and Macro and Weighted Average) and confusion matrix and have shown robust performance compared to the conventional schemes.

Development of algorithms for early detection of cyberattacks on networks using machine learning

Critical infrastructure protection and national security are enhanced by the security and reliability of networks. Various types of information circulate on these networks, ranging in classification from open to closed. The consequences of cyberattacks on these networks can be severe, including reputational damage, financial loss, operational disruption and data leakage. Traditional security methods, such as firewalls and anti-virus software, are becoming less effective against modern and ever-changing cyber threats. As a result, powerful network intrusion detection systems (IDS) have become indispensable for proactive detection and mitigation of cyber attacks. Machine learning has become a viable method for creating adaptive intrusion detection tools that can detect new and complex attack patterns. By learning from huge labelled network traffic datasets, ML models can understand the subtle patterns and differentiating features of normal and abnormal or malicious traffic flows. This allows them to detect possible cyber threats and intrusions that traditional signature-based IDSs cannot detect. Extracting discriminative features and training appropriate classification models from such data is a challenging task. In the presented study, we analyse the effectiveness of ML algorithms for detecting cyberattacks, in particular distributed denial of service (DDoS) attacks, in network traffic data. In the presented study, a network attack detection system is developed using ML and deep learning (DL) models and experimented on the CICIDS2017 dataset. The main objectives of the study are to develop a strategy for extracting valuable information from raw network streams; to study the impact of data preparation on the false positive rate; and to conduct a comparative analysis of ML models for cyberattack detection. The main goal of the study is to provide an understanding of the development of a reliable adaptive network intrusion detection system using ML approaches that increase cybersecurity capabilities and protect against future cyberattacks.

AET-net: A framework for subtype classification based on the multi-omics data of breast cancer

Breast cancer (BC) is one of the most prevalent cancers worldwide and remains a significant global public health challenge. The biomechanical characteristics of tumor microenvironments provide critical insights into cellular interactions and mechanical stress responses that potentially influence cancer progression. The integration and analysis of multi-omics data for BC subtype classification present substantial challenges, including high-dimensional data complexity and difficulties in integrating heterogeneous omics data characteristics. To address these challenges, we propose an Autoencoder and Transformer integrated neural network (AET-net) classification framework. The experimental results demonstrate that our model achieves significant performance improvements in predicting BC subtypes based on integrated multi-omics datasets, with an Accuracy of 0.912 and an AUC of 0.9862. These results not only validate the high accuracy of our model in BC subtype classification, providing a valuable tool for diagnostic decision support, but also demonstrate the potential of integrated multi-omics data analysis in enhancing the precision and efficiency of BC subtype identification.

Integrating Artificial Intelligence for Enhanced Fault Detection in Power Transmission Systems: A Smart Grid Approach

Modern electrical systems rely heavily on sensors and relays for fault detection in three-phase transmission lines and distribution transformers. However, these devices often need more time complexity and are prone to false alarms (erroneous signals). The study was guided by the PRISMA guidelines and methodology. The study's objective was to compare the levels of accuracy of fault detection presented by different models studied between 2023 and 2024. The general objective was further divided into sub-objectives, which included determining types of faults in power transmission systems, examining fault detection machine-learning models in power transmission systems, and assessing levels of accuracy of fault detection techniques in power transmission systems. The study used a qualitative research design, which entailed a systematic literature review that involved identifying scholarly articles from respectable international journal websites that aligned with examining how the deep-learning model enhanced fault detection capabilities in a three-phase transmission line simulated dataset. The inclusion criteria were that the study used journals published between 2023 and 2024. The study also included journals that tested Artificial Intelligence systems' accuracy in fault detection in three-phase transmission lines and distribution transformers. The study excluded sources that were older than 2023. The sources of information were journals written by professionals and scholars in the field of Artificial intelligence and engineering. The sourced journals used simulated and real databases to test fault detection accuracy in three-phase transmission lines and distribution transformers. The source of data for review was obtained from the internet via Google search with various credible academic journal websites being searched by the researcher. A total of 12 sources were searched. To assess the risk of bias in the included studies, the researcher used the Robvis methods that visualized risk-of-bias during the inclusion criteria for sources used in the systematic review. The techniques used to present the findings were narratives, graphs, and tables. The method used to synthesize results was comparative data analysis. The study included two studies, the sample size from the three possible journals the researcher identified. The characteristics of the studies were that they both informed the study's general and specific sub-objectives. The two studies also used the same methods to arrive at their findings, which made them ideal for comparative analysis. Finally, both studies compared older systems to new systems regarding Artificial Intelligence accuracy in fault detection. The study's findings were as follows: The study ranked the Novel glass box-based model as the best artificial intelligence machine-learning approach for accurately detecting electric power transmission lines and systems faults. The Novel glass box-based approach's accuracy was 99%, followed by the Convolution Neural Network model's accuracy of 98%, the Gated Recurrent Unit model's accuracy of 92%, the Random Forest model's accuracy of 90%, the Logistic Regression model's accuracy of 74%, and the Support Vector Classifier model's accuracy of 63%. The review was adequate to conclude that the Novel glass-box-based proposed approach was the most advanced and suitable model for detecting faults in electric power transmission lines. The study had one limitation: it reported the results based on two sources, which made the exclusion criteria a point of bias. The survey results implied that builders of new Artificial intelligence systems for fault detection in electric line transmitters should aim to produce 99% accurate systems to meet the recommendations and fill the gap that this study has reported. This study was funded by the researcher using family contributions. The study's author is Nijat Taghizad (ORCID ID: 0009-0005-0505-8767).

Prediction of soft tissue sarcoma grading using intratumoral habitats and a peritumoral radiomics nomogram: a multi-center preliminary study

BackgroundAccurate identification of pathologic grade before operation is helpful for guiding clinical treatment decisions and improving the prognosis for soft tissue sarcoma (STS).PurposeTo construct and assess a magnetic resonance imaging (MRI)-based radiomics nomogram incorporating intratumoral habitats (subregions of clusters of voxels containing similar features) and peritumoral features for the preoperative prediction of the pathological grade of STS.MethodsThe MRI data of 145 patients with STS (74 low-grade and 71 high-grade) from 4 hospitals were retrospectively collected, including enhanced T1-weighted and fat-suppressed-T2-weighted sequences. The patients were divided into training cohort (n = 102) and validation cohort (n = 43). K-means clustering was used to divide intratumoral voxels into three habitats according to signal intensity. A number of radiomics features were extracted from tumor-related regions to construct radiomics prediction signatures for seven subgroups. Logistic regression analysis identified peritumoral edema as an independent risk factor. A nomogram was created by merging the best radiomics signature with the peritumoral edema. We evaluated the performance and clinical value of the model using area under the curve (AUC), calibration curves, and decision curve analysis.ResultsA multi-layer perceptron classifier model based on intratumoral habitats and peritumoral features combined gave the best radiomics signature, with an AUC of 0.856 for the validation cohort. The AUC of the nomogram in the validation cohort was 0.868, which was superior to the radiomics signature and the clinical model established by peritumoral edema. The calibration curves and decision curve analyses revealed good calibration and a high clinical application value for this nomogram.ConclusionThe MRI-based nomogram is accurate and effective for predicting preoperative grading in patients with STS.

Sustainable Air Quality Detection Using Sequential Forward Selection-Based ML Algorithms

Air pollution has exceeded the anticipated safety limit and addressing this issue is crucial for sustainability, particularly in countries with high pollution levels. So, monitoring and forecasting air quality is essential for sustainable urban development. Therefore, this paper presents multiclass classification using two feature selection techniques, namely Sequential Forward Selection (SFS) and filtering, both with different machine learning and ensemble techniques, to predict air quality and make sure that the most relevant features are included in datasets for air quality determination. The results of the considered framework reveal that the SFS technique provides superior performance compared to filter feature selection (FFS) with different ML methods, including the AdaBoost Classifier, the Extra Tree Classifier, Random Forest (RF), and the Bagging Classifier, for efficiently determining the Air Quality Index (AQI). These models’ performances are assessed using predetermined performance metrics. The AdaBoost Classifier model with FFS has the lowest accuracy, while the RF model with SFS achieves the highest accuracy, at 78.4% and 99.99%, respectively. Based on the raw dataset, it was noted that the F1-score, recall, and precision values of the RF model with SFS are 99.96%, 99.97%, and 99.98%, respectively. Therefore, the experimental results undoubtedly show the supremacy, reliability, and robustness of the proposed approach in determining the AQI effectively.

Detecting Online Sexism: Integrating Sentiment Analysis with Contextual Language Models

In the digital era, social media platforms have seen a substantial increase in the volume of online comments. While these platforms provide users with a space to express their opinions, they also serve as fertile ground for the proliferation of hate speech. Hate comments can be categorized into various types, including discrimination, violence, racism, and sexism, all of which can negatively impact mental health. Among these, sexism poses a significant challenge due to its various forms and the difficulty in defining it, making detection complex. Nevertheless, detecting and preventing sexism on social networks remains a critical issue. Recent studies have leveraged language models such as transformers, known for their ability to capture the semantic nuances of textual data. In this study, we explore different transformer models, including multiple versions of RoBERTa (A Robustly Optimized BERT Pretraining Approach), to detect sexism. We hypothesize that combining a sentiment-focused language model with models specialized in sexism detection can improve overall performance. To test this hypothesis, we developed two approaches. The first involved using classical transformers trained on our dataset, while the second combined embeddings generated by transformers with a Long Short-Term Memory (LSTM) model for classification. The probabilistic outputs of each approach were aggregated through various voting strategies to enhance detection accuracy. The LSTM with embeddings approach improved the F1-score by 0.2% compared to the classical transformer approach. Furthermore, the combination of both approaches confirms our hypothesis, achieving a 1.6% improvement in the F1-score in each case. We determined that an F1 score of over 0.84 effectively measures sexism. Additionally, we constructed our own dataset to train and evaluate the models.

Assessment of image quality on the diagnostic performance of clinicians and deep learning models: Cross-sectional comparative reader study.

Skin cancer is a prevalent and clinically significant condition, with early and accurate diagnosis being crucial for improved patient outcomes. Dermoscopy and artificial intelligence (AI) hold promise in enhancing diagnostic accuracy. However, the impact of image quality, particularly high dynamic range (HDR) conversion in smartphone images, on diagnostic performance remains poorly understood. This study aimed to investigate the effect of varying image qualities, including HDR-enhanced dermoscopic images, on the diagnostic capabilities of clinicians and a convolutional neural network (CNN) model. Eighteen dermatology clinicians assessed 303 images of 101 skin lesions that were categorized into three image quality groups: low quality (LQ), high quality (HQ) and enhanced quality (EQ) produced using HDR-style conversion. Clinicians participated in a two part reader study that required their diagnosis, management and confidence level for each image assessed. In the binary classification of lesions, clinicians had the greatest diagnostic performance with HQ images, with sensitivity (77.3%; CI 69.1-85.5), specificity (63.1%; CI 53.7-72.5) and accuracy (70.2%; CI 61.3-79.1). For the multiclass classification, the overall performance was also best with HQ images, attaining the greatest specificity (91.9%; CI 83.2-95.0) and accuracy (51.5%; CI 48.4-54.7). Clinicians had a superior performance (median correct diagnoses) to the CNN model for the binary classification of LQ and EQ images, but their performance was comparable on the HQ images. However, in the multiclass classification, the CNN model significantly outperformed the clinicians on HQ images (p < 0.01). This study highlights the importance of image quality on the diagnostic performance of clinicians and deep learning models. This has significant implications for telehealth reporting and triage.

An efficient unsupervised classification model for galaxy morphology: Voting clustering based on coding from ConvNeXt large model

By combining unsupervised and supervised machine learning methods, we have proposed a framework, called USmorph to carry out automatic classifications of galaxy morphologies. In this work, we update the unsupervised machine learning (UML) step by proposing an algorithm based on ConvNeXt large model coding to improve the efficiency of unlabeled galaxy morphology classifications. The method can be summarized into three key aspects as follows: (1) a convolutional autoencoder is used for image denoising and reconstruction and the rotational invariance of the model is improved by polar coordinate extension; (2) utilizing a pre-trained convolutional neural network (CNN) named ConvNeXt for encoding the image data. The features were further compressed via a principal component analysis (PCA) dimensionality reduction; (3) adopting a bagging-based multi-model voting classification algorithm to enhance robustness. We applied this model to I-band images of a galaxy sample with $I_ mag &lt; 25$ in the COSMOS field. Compared to the original unsupervised method, the number of clustering groups required by the new method is reduced from 100 to 20. Finally, we managed to classify about 53&lt;!PCT!&gt; galaxies, significantly improving the classification efficiency. To verify the validity of the morphological classification, we selected massive galaxies with $M_ M_ sun $ for morphological parameter tests. The corresponding rules between the classification results and the physical properties of galaxies on multiple parameter surfaces are consistent with the existing evolution model. Our method has demonstrated the feasibility of using large model encoding to classify galaxy morphology, which not only improves the efficiency of galaxy morphology classification, but also saves time and manpower. Furthermore, in comparison to the original UML model, the enhanced classification performance is more evident in qualitative analysis and has successfully surpassed a greater number of parameter tests. The enhanced UML method will support the Chinese space station telescope in the future.

Multivariate analysis of socioeconomic factors in SMEs: regression models machine learning

The application of multivariate analysis in various social and economic sectors allows the implementation and denoting of the virtues that this statistical method has and the relevance that is generated in the presentation of results for the decision-making necessary in the development of SMEs, allowing to indicate the various factors that lead to their growth and consolidation in the market. The factors that have a significant impact on these types of companies are analyzed, through the application of multivariate techniques such as Factor Analysis, Principal Components, Cluster, and Classification Models that in turn have allowed a reduction from 211 variables to 32 latent variables that focus the initial information, obtaining that the opportunities and capabilities, physical environment and health and job satisfaction that constitute SMEs are the factors that have the greatest influence have in their consolidation.

Hyperspectral Imaging for Detecting Plastic Debris on Shoreline Sands to Support Recycling

Environmental pollution from plastic debris is raising concerns not only for the vulnerability of marine species to ingestion but also for potential human health hazards posed by small particles, known as microplastics. In this context, marine areas suffer from a lack of constant shoreline cleanups to remove accumulated debris, preventing their degradation and fragmentation. To establish optimal strategies for streamlining plastic recovery and recycling operations, it is important to have a system for recognizing plastic debris on the beach and, more specifically, for identifying the type of polymer and mapping (e.g., topologically assessing) the distribution of plastic debris on shoreline sands. This study aims to provide an operative tool finalized to perform an in situ detection, analysis, and characterization of plastic debris present in the coastal environment (i.e., beaches), adopting a near-infrared (NIR)-based hyperspectral imaging (HSI) approach. In more detail, the possibility of identifying and classifying polymers of plastic debris by NIR-HSI in three different areas along the Pontine coastline of the Lazio region (Latina, Italy) was investigated. The study focused on three distinct beaches (i.e., Foce Verde, Capo Portiere, and Sabaudia), each characterized by a different type of sand. For each location, the adopted approach allowed for the systematic classification of the various types of plastic waste found. Three Partial Least Squares Discriminant Analysis (PLS-DA) classification models were developed using a cascade detection strategy. The first model was designed to distinguish plastics from other materials in sand samples, the second to detect plastic particles in the sand, and the third to classify the type of polymer composing each identified plastic particle. Obtained results showed that, on the one hand, plastics were correctly detected from sand and other materials (i.e., sensitivity = 0.892–1.000 and specificity = 0.909–0.996), and on the other, the recognition of polymer type was satisfactory, according to the performance statistical parameters (i.e., sensitivity = 1.000 and specificity = 0.991–1.000). This research highlights the potential of the NIR-HSI approach as a reliable, non-invasive method for plastic debris monitoring and polymer classification. Its scalability and adaptability suggest possible future integration into mobile systems, enabling large-scale monitoring and efficient debris management.

How transfer learning is used in generative models for image classification: improved accuracy

How transfer learning is used in generative models for image classification: improved accuracy

Breast cancer detection and classification with digital breast tomosynthesis: a two-stage deep learning approach.

The purpose of this study was to propose a new computer-assisted two-staged diagnosis system that combines a modified deep learning (DL) architecture (VGG19) for the classification of digital breast tomosynthesis (DBT) images with the detection of tumors as benign or cancerous using the You Only Look Once version 5 (YOLOv5) model combined with the convolutional block attention module (CBAM) (known as YOLOv5-CBAM). In the modified version of VGG19, eight additional layers were integrated, comprising four batch normalization layers and four additional pooling layers (two max pooling and two average pooling). The CBAM was incorporated into the YOLOv5 model structure after each feature fusion. The experiment was carried out using a sizable benchmark dataset of breast tomography images. A total of 22,032 DBT examinations from 5,060 patients were included in the data. Test accuracy, training loss, and training accuracy showed better performance with our proposed architecture than with previous models. Hence, the modified VGG19 classified DBT images more accurately than previously possible using pre-trained model-based architectures. Furthermore, a YOLOv5-based CBAM precisely discriminated between benign lesions and those that were malignant. DBT images can be classified using modified VGG19 with accuracy greater than the previously available pre-trained models-based architectures. Furthermore, a YOLOv5-based CBAM can precisely distinguish between benign and cancerous lesions. The proposed two-tier DL algorithm, combining a modified VGG19 model for image classification and YOLOv5-CBAM for lesion detection, can improve the accuracy, efficiency, and reliability of breast cancer screening and diagnosis through innovative artificial intelligence-driven methodologies.

Systematic Review of Generative Modelling Tools and Utility Metrics for Fully Synthetic Tabular Data

Sharing data with third parties is essential for advancing science, but it is becoming more and more difficult with the rise of data protection regulations, ethical restrictions, and growing fear of misuse. Fully synthetic data, which transcends anonymisation, may be the key to unlocking valuable untapped insights stored away in secured data vaults. This review examines current synthetic data generation methods and their utility measurement. We found that more traditional generative models such as Classification and Regression Tree models alongside Bayesian Networks remain highly relevant and are still capable of surpassing deep learning alternatives like Generative Adversarial Networks. However, our findings also display the same lack of agreement on metrics for evaluation, uncovered in earlier reviews, posing a persistent obstacle to advancing the field. We propose a tool for evaluating the utility of synthetic data and illustrate how it can be applied to three synthetic data generation models. By streamlining evaluation and promoting agreement on metrics, researchers can explore novel methods and generate compelling results that will convince data curators and lawmakers to embrace synthetic data. Our review emphasises the potential of synthetic data and highlights the need for greater collaboration and standardisation to unlock its full potential.

Spot the bot: the inverse problems of NLP

This article concerns the problem of distinguishing human-written and bot-generated texts. In contrast to the classical problem formulation, in which the focus falls on one type of bot only, we consider the problem of distinguishing texts written by any person from those generated by any bot; this involves analysing the large-scale, coarse-grained structure of the language semantic space. To construct the training and test datasets, we propose to separate not the texts of bots, but bots themselves, so the test sample contains the texts of those bots (and people) that were not in the training sample. We aim to find efficient and versatile features, rather than a complex classification model architecture that only deals with a particular type of bots. In the study we derive features for human-written and bot generated texts, using clustering (Wishart and K-Means, as well as fuzzy variations) and nonlinear dynamic techniques (entropy-complexity measures). We then deliberately use the simplest of classifiers (support vector machine, decision tree, random forest) and the derived characteristics to identify whether the text is human-written or not. The large-scale simulation shows good classification results (a classification quality of over 96%), although varying for languages of different language families.

Does Using Artificial Intelligence in Citizen Science Support Volunteers&amp;rsquo; Learning? An Experimental Study in Ornithology

One of the oldest and largest biodiversity-related citizen science (CS) projects is eBird (https://ebird.org/home), developed by the Cornell Lab of Ornithology. It provides a mobile application for birdwatchers to record checklists of when, where, and how they have seen or heard birds. The Cornell Lab has also developed a mobile application, Merlin, which uses a deep convolutional neural network to help users automatically identify bird species from photos, sounds (converted to spectrograms), or descriptions. This research investigates how the use of machine learning (ML) classification models affects the learning of novice birders. Our participants (computer science students with no previous background in ornithology) were randomly divided into three groups: one using the eBird application and identifying bird species themselves; one using the Merlin application, which uses ML to automatically identify birds from photos or sounds; and a control group. Participants were tested on their knowledge of birds before and after participating in the project to see how using the ML classification model affected their learning. We also interviewed selected participants after the post-test to understand what they had done and what might explain the results. Our results show that novice participants who participate in a CS project for even a short time significantly improve their content knowledge of familiar birds in their neighbourhood, and that eBird users outperform Merlin users on the knowledge post-test. Although AI may improve volunteer productivity and retention, there is a risk that it may reduce their learning. Further research with different participant profiles and project designs is needed to understand how to optimise volunteer productivity, retention, and learning in AI-assisted CS projects.