Search

Feature selection in Knowledge Graphs (KGs) is increasingly utilized in diverse domains, including biomedical research, Natural Language Processing (NLP), and personalized recommendation systems. This paper delves into the methodologies for feature selection (FS) within KGs, emphasizing their roles in enhancing machine learning (ML) model efficacy, hypothesis generation, and interpretability. Through this comprehensive review, we aim to catalyze further innovation in FS for KGs, paving the way for more insightful, efficient, and interpretable analytical models across various domains. Our exploration reveals the critical importance of scalability, accuracy, and interpretability in FS techniques, advocating for the integration of domain knowledge to refine the selection process. We highlight the burgeoning potential of multi-objective optimization and interdisciplinary collaboration in advancing KG FS, underscoring the transformative impact of such methodologies on precision medicine, among other fields. The paper concludes by charting future directions, including the development of scalable, dynamic FS algorithms and the integration of explainable AI principles to foster transparency and trust in KG-driven models.

ObjectiveRetinal prostheses evoke visual precepts by electrically stimulating functioning cells in the retina. Despite high variance in perceptual thresholds across subjects, among electrodes within a subject, and over time, retinal prosthesis users must undergo `system fitting', a process performed to calibrate stimulation parameters according to the subject's perceptual thresholds. Although previous work has identified electrode-retina distance and impedance as key factors affecting thresholds, an accurate predictive model is still lacking.ApproachTo address these challenges, we 1) fitted machine learning (ML) models to a large longitudinal dataset with the goal of predicting individual electrode thresholds and deactivation as a function of stimulus, electrode, and clinical parameters (`predictors') and 2) leveraged explainable artificial intelligence (XAI) to reveal which of these predictors were most important.Main resultsOur models accounted for up to 76% of the perceptual threshold response variance and enabled predictions of whether an electrode was deactivated in a given trial with F1 and AUC scores of up to 0.730 and 0.910, respectively. Our models identified novel predictors of perceptual sensitivity, including subject age, time since blindness onset, and electrode-fovea distance.SignificanceOur results demonstrate that routinely collected clinical measures and a single session of system fitting might be sufficient to inform an XAI-based threshold prediction strategy, which has the potential to transform clinical practice in predicting visual outcomes..

This study aimed to investigate various models for predicting the cutting force in rock-cutting processes by conical tools or pick cutters. For this purpose, a database of rock cutting forces was established by utilizing full-scale cutting tests and analysis of the measured forces as a function of input parameters, such as uniaxial compressive strength (UCS), Brazilian tensile strength (BTS), spacing, and the contact area between the pick cutter and rock. The study outlined the procedure for calculating the area of contact (AOC) between a conical pick cutter and rock surface, utilizing key parameters, including tip radius, tip cone angle, depth of penetration, and a fixed 45° attack angle. Six categories of regression models were employed, encompassing conventional regression models (linear, log–log linear, and polynomial), regularized models (LASSO, Ridge, and Elastic-Net), tree-based models (decision tree, random forest, and extreme gradient boost), complex models (SVR and ANN), probabilistic (Bayesian linear and Gaussian process), and Ensemble models (stacking and voting). As a result, the stacking technique within the ensemble models exhibited superior performance in predicting cutting forces, showing the highest Coefficient of Determination (R2) score and the lowest Mean Absolute Error (MAE). To enhance the interpretability of the results, particularly from the ensemble methods, Explainable Artificial Intelligence (XAI) techniques, such as individual conditional expectation (ICE), partial dependence plots (PDPs), and SHAP (SHapley Additive exPlanations) analysis, were applied. The research offers reasonable prediction of cutting force based on specified parameters. This empowers engineers to make well-informed decisions regarding cutter selection, machine specifications, and cutting strategies, resulting in more efficient rock-cutting operations. Highlights: The study introduced the contact area between the pick cutter and the rock surface as a calculated input parameter for predicting cutting force. Six categories of regression models were used to predict cutting force, and their performances were compared with each other. The results highlight the efficacy of ensemble techniques, particularly the stacking model, in accurately predicting cutting force. [ABSTRACT FROM AUTHOR]

Chronic Obstructive Pulmonary Disease (COPD) is a predominant global health concern, ranking third in mortality rates, yet frequently remains undiagnosed until its advanced stages. Given its prevalence, the need for innovative and widely accessible diagnostic tools has never been more paramount. While spirometry tests serve as conventional diagnostic benchmarks, their reach remains limited, especially in regions with constrained medical resources. The presented research harnesses deep learning algorithms to facilitate early-stage COPD detection, specifically targeting Chest X-rays (CXRs). The clinically annotated VinDR-CXR dataset provides the primary foundation for model training, complemented by incorporating the ChestX-ray14 dataset for initial model pre-training. Such a dualdataset strategy augments model generalization and adaptability. Among several explored Convolutional Neural Network (CNN) architectures, the Xception model emerges as a frontrunner. Through transfer learning methodologies, this model produces a noteworthy recall rate of 98.2%, markedly surpassing the metrics of the ResNet50 model. Recognizing the imperative for transparency in AI applications in medical imaging, the research integrates essential explainability approaches viz: Gradient Class Activation Mapping (Grad-CAM) and SHapley Additive exPlanations (SHAP). These techniques elucidate the AI's decision-making process, offering invaluable visual and analytical insights for fostering trust among medical professionals. In essence, this study not only underscores the potential of integrating AI with medical imaging for COPD detection but also accentuates the pivotal role of transparency in AI-driven medical interventions. [ABSTRACT FROM AUTHOR]

In order to maintain sustainable agriculture, it is vital to monitor plant health. Since all species of plants are prone to characteristic diseases, it necessitates regular surveillance to search for any symptoms, which is utterly challenging and time-consuming. Besides, farmers may struggle to identify the type of plant disease and its potential symptoms. Hence, the interest in research like image-based computer-aided automated plant leaf disease detection by analyzing the early symptoms has increased enormously. However, limitations in the plant leaf image database, for instance, unfitting backgrounds, blurry images, and so on, sometimes cause underprivileged feature extraction, misclassification, and overfitting issues in existing models. As a result, we have proposed a real-time plant leaf disease detection architecture incorporating proposed PLD-Det model, which is based on improved YOLOv7 with the intention of assisting farmers while reducing the issues in existing models. The architecture has been trained on the widely used PlantVillage dataset, which resulted in an accuracy of 98.53%. Furthermore, SHapley Additive exPlanations (SHAP) values have been analyzed as a unified measure of feature significance. According to the experimental findings, the proposed PLD-Det model, which is an improved YOLOv7 architecture, outperformed the original YOLOv7 model in test accuracy by approximately 4%. [ABSTRACT FROM AUTHOR]

Forest fires pose a serious risk to ecosystems in the Mediterranean region; thus, 2021 fires in the Mediterranean region of Türkiye emphasize the requirement for accurate and interpretable forest fire susceptibility (FFS) mapping. This study presents an innovative approach to FFS mapping for the Mersin, Antalya, and Mugla provinces, integrating machine learning (ML) models with Explainable Artificial Intelligence (XAI). The methodology employs three state-of-the-art ML models: eXtreme Gradient Boosting (XGBoost), Gradient Boosting Machine (GBM), and Light Gradient-Boosting Machine (LightGBM). These models generated FFS maps using 14 fire conditioning factors, including meteorological, topographic, environmental, and anthropogenic factors. LightGBM demonstrated outstanding performance, acquiring the highest accuracy (0.897), outperforming GBM (0.881) and XGBoost (0.851). McNemar's statistical test demonstrated significant differences in the predictive capabilities between XGBoost and both GBM and LightGBM, whereas no significant difference was found between GBM and LightGBM. Information Gain and SHapley Additive exPlanations (SHAP) analyses were applied to enhance model interpretability and validate feature importance. Both methods agreed that the most influential variables in FFS are soil moisture, Palmer Drought Severity Index (PDSI), and Land Surface Temperature (LST). On the other hand, SHAP plots revealed complex, nonlinear relationships between these factors and fire susceptibility. At the same time, a high increase in LST enhances the risk of fires; higher soil moisture values and the PDSI decrease the possibility of fire risk. This research also contributes to the concept of FFS mapping interpretability and operational utility with the application of XAI, which establishes a transparent basis for identifying fire risk drivers in Mediterranean ecosystems. [ABSTRACT FROM AUTHOR]

One of the major obstacles to the widespread adoption of Virtual Reality (VR) is Cybersickness. It is a sense of physical discomfort akin to motion sickness experienced by the users either during or subsequent to VR utilization. Typically, it is detected through explicit methods such as self-reported questionnaires, which are not ideal for online and implicit monitoring of users’ well-being. This study tackles the challenge of implicitly detecting and measuring Cybersickness in VR through physiological signals. Therefore, we propose a multimodal approach that integrates physiological signals with self-reported measures. We utilize a mixed-methods design combining quantitative and qualitative analyses, using a roller coaster simulation as the experimental paradigm. The research analyzes physiological data collected from a group of 22 participants exposed to a simulated VR rollercoaster experience, through statistical analysis and machine learning algorithms. The physiological markers studied include Electroencephalograms (EEG), Electrodermal Activity, Blood Volume Pulse, and skin Temperature signals. Additionally, the study elucidates the complex interplay among physiological markers using Explainable AI (XAI). We found out significant correlations between high Simulator Sickness Questionnaire scores and physiological measures such as brain rhythms and EEG indices related to engagement, visual fatigue, and drowsiness, as well as heart rate variability. Moreover, the proposed machine learning model achieves an accuracy of 86.66 % in detecting elevated Cybersickness symptoms, which is higher than the accuracy achieved by existing techniques. Further exploration using the XAI technique confirms the elevated levels of drowsiness and reduced engagement in participants experiencing elevated Cybersickness. By providing a comprehensive, multimodal approach for quantitative assessment, this study fills a gap in the existing literature and paves the path for the development of adaptive systems that modulate their behavior based on physiological data. [ABSTRACT FROM AUTHOR]

In quest of improving the productivity and efficiency of manufacturing processes, Artificial Intelligence (AI) is being used extensively for response prediction, model dimensionality reduction, process optimization, and monitoring. Though having superior accuracy, AI predictions are unintelligible to the end users and stakeholders due to their opaqueness. Thus, building interpretable and inclusive machine learning (ML) models is a vital part of the smart manufacturing paradigm to establish traceability and repeatability. The study addresses this fundamental limitation of AI-driven manufacturing processes by introducing a novel Explainable AI (XAI) approach to develop interpretable processes and product fingerprints. Here the explainability is implemented in two stages: by developing interpretable representations for the fingerprints, and by posthoc explanations. Also, for the first time, the concept of process fingerprints is extended to develop an interpretable probabilistic model for bottleneck events during manufacturing processes. The approach is demonstrated using two datasets: nanosecond pulsed laser ablation to produce superhydrophobic surfaces and wire EDM real-time monitoring dataset during the machining of Inconel 718. The fingerprint identification is performed using a global Lipschitz functions optimization tool (MaxLIPO) and a stacked ensemble model is used for response prediction. The proposed interpretable fingerprint approach is robust to change in processes and can responsively handle both continuous and categorical responses alike. Implementation of XAI not only provided useful insights into the process physics but also revealed the decision-making logic for local predictions. [ABSTRACT FROM AUTHOR]

Anomaly detection in medical imaging, particularly within the realm of magnetic resonance imaging (MRI), stands as a vital area of research with far‐reaching implications across various medical fields. This review meticulously examines the integration of artificial intelligence (AI) in anomaly detection for MR images, spotlighting its transformative impact on medical diagnostics. We delve into the forefront of AI applications in MRI, exploring advanced machine learning (ML) and deep learning (DL) methodologies that are pivotal in enhancing the precision of diagnostic processes. The review provides a detailed analysis of preprocessing, feature extraction, classification, and segmentation techniques, alongside a comprehensive evaluation of commonly used metrics. Further, this paper explores the latest developments in ensemble methods and explainable AI, offering insights into future directions and potential breakthroughs. This review synthesizes current insights, offering a valuable guide for researchers, clinicians, and medical imaging experts. It highlights AI's crucial role in improving the precision and speed of detecting key structural and functional irregularities in MRI. Our exploration of innovative techniques and trends furthers MRI technology development, aiming to refine diagnostics, tailor treatments, and elevate patient care outcomes. Level of Evidence: 5 Technical Efficacy: Stage 1. [ABSTRACT FROM AUTHOR]

As the use of drones continues to increase, their capabilities pose a threat to airspace safety when they are misused. Deploying AI models for intercepting these unwanted drones becomes crucial. However, these AI models, such as deep learning models, often operate as "black boxes", making it hard to trust their decision-making system. This also affects end-users' confidence in these AI systems. In this paper, the explainability of deep reinforcement learning is investigated and a deep reinforcement learning (DRL) method, double deep Q-network with dueling architecture and prioritized experience replay is applied to train the AI models. To make the AI model decisions more transparent and to understand the reasoning behind the AI decisions for counter-drone systems, Shapley Additive Explanations (SHAP) method is implemented. After training the DRL agent, experience replay is visualized, and the absolute SHAP values are calculated to explain the key factors that influence the deep reinforcement learning agent's choices. The integration of DRL with explainable AI methods such as SHAP demonstrates significant potential for the advancement of robust and efficient counter-drone systems. [ABSTRACT FROM AUTHOR]

The Covid-19 disease has spread widely over the whole world since the beginning of 2020. Following the epidemic which started in Wuhan, China on January 30, 2020 the World Health Organization (WHO) declared a global health emergency and a pandemic. Researchers of different disciplines work along with public health officials to understand the SARS-CoV-2 pathogenesis and jointly with the policymakers urgently develop strategies to control the spread of this new disease. Recent findings have observed specific image patterns from computed tomography (CT) for patients infected by SARS-CoV-2 which are distinct from the other pulmonary diseases. In this paper, we propose an explainable-by-design that has an integrated image segmentation mechanism based on SLIC that improves the algorithm performance and the interpretability of the resulting model. In order to evaluate the proposed approach, we used the SARS-CoV-2 CT scan dataset that we published recently and has been widely used in the literature. The proposed Super-xDNN could obtain statistically better results than traditional deep learning approaches as DenseNet-201 and Resnet-152. Furthermore, it also improved the explainability and interpretability of its decision mechanism when compared with the xDNN basis approach that uses the whole image as prototype. The segmentation mechanism of Super-xDNN favored a decision structure that is more close to the human logic. Moreover, it also allowed the provision of new insights as a heat-map which highlights the areas with highest similarities with Covid-19 prototypes, and an estimation of the area affected by the disease. [ABSTRACT FROM AUTHOR]

This study leverages deep learning to analyze sex differences in brain MRI data, aiming to further advance fairness in medical imaging. We employed 3D T1-weighted Magnetic Resonance images from four diverse datasets: Calgary-Campinas-359, OASIS-3, Alzheimer's Disease Neuroimaging Initiative, and Cambridge Center for Aging and Neuroscience, ensuring a balanced representation of sexes and a broad demographic scope. Our methodology focused on minimal preprocessing to preserve the integrity of brain structures, utilizing a Convolutional Neural Network model for sex classification. The model achieved an accuracy of 87% on the test set without employing total intracranial volume (TIV) adjustment techniques. We observed that while the model exhibited biases at extreme brain sizes, it performed with less bias when the TIV distributions overlapped more. Saliency maps were used to identify brain regions significant in sex differentiation, revealing that certain supratentorial and infratentorial regions were important for predictions. Furthermore, our interdisciplinary team, comprising machine learning specialists and a radiologist, ensured diverse perspectives in validating the results. The detailed investigation of sex differences in brain MRI in this study, highlighted by the sex differences map, offers valuable insights into sex-specific aspects of medical imaging and could aid in developing sex-based bias mitigation strategies, contributing to the future development of fair AI algorithms. Awareness of the brain's differences between sexes enables more equitable AI predictions, promoting fairness in healthcare outcomes. Our code and saliency maps are available at https://github.com/mahsadibaji/sex-differences-brain-dl. [ABSTRACT FROM AUTHOR]

AI is anticipated to improve human decision-making across various domains, often in high-stakes, difficult tasks. However, human reliance on AI recommendations is often inappropriate. A common approach to address this is to provide explanations about the AI output to decision makers, but results have been mixed so far. It often remains unclear when people can rely appropriately on AI and when explanations can help. In this work, we conducted a lab experiment (N = 34) to investigate how the appropriateness of human reliance on (explainable) AI depends on the mental workload induced by different decision difficulties. Instead of self-assessments, we used EEG (Emotiv Epoc Flex head cap, 32 wet electrodes) to more directly measure participants' mental workload. We found that the difficulty of a decision, indicated by the induced mental workload, strongly influences participants' ability to rely appropriately on AI, as assessed through relative self-reliance, relative AI reliance, and decision accuracy with and without AI. While reliance was appropriate for low mental workload decisions, participants were prone to overreliance in high mental workload decisions. Explanations had no significant effect in either case. Our results imply that alternatives to the common ‘recommend-and-explain’ approach should be explored to assist human decision-making in challenging tasks. [ABSTRACT FROM AUTHOR]

This study provides a comprehensive bibliometric and in-depth analysis of artificial intelligence (AI) and machine learning (ML) applications in environmental monitoring, based on 4762 publications from 1991 to 2024. The research highlights a notable increase in publications and citations since 2010, with China, the United States, and India emerging as leading contributors. Key areas of research include air and water quality monitoring, climate change modeling, biodiversity assessment, and disaster management. The integration of AI with emerging technologies, such as the Internet of Things (IoT) and remote sensing, has significantly expanded real-time environmental monitoring capabilities and data-driven decision-making. In-depth analysis reveals advancements in AI/ML methodologies, including novel algorithms for soil mapping, land-cover classification, flood susceptibility modeling, and remote sensing image analysis. Notable applications include enhanced air quality predictions, water quality assessments, climate impact forecasting, and automated wildlife monitoring using AI-driven image recognition. Challenges such as the "black-box" nature of AI models, the need for high-quality data in resource-constrained regions, and the complexity of real-time disaster management are also addressed. The study highlights ongoing efforts to develop explainable AI (XAI) models, which aim to improve model transparency and trust in critical environmental applications. Future research directions emphasize improving data quality and availability, fostering interdisciplinary collaborations across environmental and computer sciences, and addressing ethical considerations in AI-driven environmental management. These findings underscore the transformative potential of AI and ML technologies for sustainable environmental management, offering valuable insights for researchers and policymakers in addressing global environmental challenges. [ABSTRACT FROM AUTHOR]

Background/Objectives: Stroke stands as a prominent global health issue, causing con-siderable mortality and debilitation. It arises when cerebral blood flow is compromised, leading to irreversible brain cell damage or death. Leveraging the power of machine learning, this paper presents a systematic approach to predict stroke patient survival based on a comprehensive set of factors. These factors include demographic attributes, medical history, lifestyle elements, and physiological metrics. Method: An effective random sampling method is proposed to handle the highly biased data of stroke. The stroke pre-diction using optimized boosting machine learning algorithms is supported with explainable AI using LIME and SHAP. This enables the models to discern intricate data patterns and establish correlations between selected features and patient survival. Results: The performance of three boosting algorithms is studied for stroke prediction, which include Gradient Boosting (GB), AdaBoost (ADB), and XGBoost (XGB) with XGB achieved the best outcome overall with a training accuracy of 96.97% and testing accuracy of 92.13%. Conclusions: Through this approach, the study seeks to uncover actionable insights to guide healthcare practitioners in devising personalized treatment strategies for stroke patients. [ABSTRACT FROM AUTHOR]

SUMMARY: Plant leaves play a pivotal role in automated species identification using deep learning (DL). However, achieving reproducible capture of leaf variation remains challenging due to the inherent "black box" problem of DL models. To evaluate the effectiveness of DL in capturing leaf shape, we used geometric morphometrics (GM), an emerging component of eXplainable Artificial Intelligence (XAI) toolkits. We photographed Ranunculus auricomus leaves directly in situ and after herbarization. From these corresponding leaf images, we automatically extracted DL features using a neural network and digitized leaf shapes using GM. The association between the extracted DL features and GM shapes was then evaluated using dimension reduction and covariation models. DL features facilitated the clustering of leaf images by source populations in both in situ and herbarized leaf image datasets, and certain DL features were significantly associated with biological leaf shape variation as inferred by GM. DL features also enabled leaf classification into morpho‐phylogenomic groups within the intricate R. auricomus species complex. We demonstrated that simple in situ leaf imaging and DL reproducibly captured leaf shape variation at the population level, while combining this approach with GM provided key insights into the shape information extracted from images by computer vision, a necessary prerequisite for reliable automated plant phenotyping. Significance Statement: By integrating geometric morphometrics and explainable artificial intelligence, we revealed significant correlations between computer vision perception of images of plant leaves and the true leaf shapes captured in the images. We demonstrated that deep learning achieves fine‐scale leaf shape recognition that differentiates even populations within a complicated plant species complex, highlighting its potential for reproducible and interpretable automated plant phenotyping. [ABSTRACT FROM AUTHOR]

Alzheimer's disease (AD) is the most common cause of dementia, marked by cognitive decline and memory loss. Recently, machine learning and deep learning techniques have introduced promising solutions for improving AD detection through MRI, especially in settings where specialists may not be readily available. These techniques offer the potential to assist general practitioners and non-specialists in busy clinical environments. However, the 'black box' nature of many AI techniques makes it challenging for non-expert physicians to fully trust their diagnostic accuracy. In this review, we critically evaluate current explainable AI (XAI) methods applied to AD detection and highlight their limitations. In addition, a new interpretability framework, called "Feature-Augmented", was theoretically designed to improve model interpretability. This approach remains underexplored, primarily due to the scarcity of explainable AD-specific datasets. Furthermore, we underscore the importance of AI models being accurate and explainable, which enhance diagnostic confidence and patient care outcomes. [ABSTRACT FROM AUTHOR]

The surface defect has a direct impact on the performance and quality of the final product, so it is crucial to identify the key factors causing the defect. To achieve accurate key factor identification, an eXplainable AI‐based Pipeline (XAIP) is proposed herein. The proposed XAIP combines data mode clustering and local model construction to meet practical application requirements. Meanwhile, after analyzing the characteristics of practical data, a strategy for defect rate calculation is designed to provide more fine‐grained defect‐related information to supervise model training. The final identification results are presented by a two‐stage explainable method, which is designed to reveal the information of data modes and key variables and can give engineers more specific defect‐related information. The experiment results based on two practical manufacturing datasets show the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Mobile application developers rely largely on user reviews for identifying issues in mobile applications and meeting the users' expectations. User reviews are unstructured, unorganized and very informal. Identifying and classifying issues by extracting required information from reviews is difficult due to a large number of reviews. To automate the process of classifying reviews many researchers have adopted machine learning approaches. Keeping in view, the rising demand for educational applications, especially during COVID-19, this research aims to automate Android application education reviews' classification and sentiment analysis using natural language processing and machine learning techniques. A baseline corpus comprising 13,000 records has been built by collecting reviews of more than 20 educational applications. The reviews were then manually labelled with respect to sentiment and issue types mentioned in each review. User reviews are classified into eight categories and various machine learning algorithms are applied to classify users' sentiments and issues of applications. The results demonstrate that our proposed framework achieved an accuracy of 97% for sentiment identification and an accuracy of 94% in classifying the most significant issues. Moreover, the interpretability of the model is verified by using the explainable artificial intelligence technique of local interpretable model-agnostic explanations. [ABSTRACT FROM AUTHOR]

Explainability in the field of event detection is a new emerging research area. For practitioners and users alike, explainability is essential to ensuring that models are widely adopted and trusted. Several research efforts have focused on the efficacy and efficiency of event detection. However, a human‐centric explanation approach to existing event detection solutions is still lacking. This paper presents an overview of a conceptual framework for human‐centric semantic‐based explainable event detection with the acronym HUSEED. The framework considered the affordances of XAI and semantics technologies for human‐comprehensible explanations of events to facilitate 5W1H explanations (Who did what, when, where, why, and how). Providing this kind of explanation will lead to trustworthy, unambiguous, and transparent event detection models with a higher possibility of uptake by users in various domains of application. We illustrated the applicability of the proposed framework by using two use cases involving first story detection and fake news detection. [ABSTRACT FROM AUTHOR]

We present an exciting journey from 3D point‐cloud data (PCD) to the state of the art in graph neural networks (GNNs) and their evolution with explainable artificial intelligence (XAI), and 3D geometric priors with the human‐in‐the‐loop. We follow a simple definition of a "digital twin," as a high‐precision, three‐dimensional digital representation of a physical object or environment, captured, for example, by Light Detection and Ranging (LiDAR) technology. After a digression into transforming PCD into images, graphs, combinatorial complexes and hypergraphs, we explore recent developments in geometric deep learning (GDL) and provide insight into the application of these network architectures for analyzing and learning from graph‐structured data. We emphasize the importance of the explainability of these models and recognize that the ability to interpret and validate the results of complex models is a crucial aspect of their wider adoption. This article is categorized under:Technologies > Artificial Intelligence [ABSTRACT FROM AUTHOR]

Anomaly detection is an important feature in modern additive manufacturing (AM) systems to ensure quality of the produced components. Although this topic is well discussed in the literature, current methods rely on black-box approaches, limiting our understanding of why anomalies occur, making complex the root cause identification and the consequent decision support about the action to take to mitigate them. This work addresses these limitations by proposing a structured workflow designed to enhance the explainability of anomaly detection models. Using the wire arc additive manufacturing (WAAM) process as a case study, we examined 14 wall structures printed with INVAR36 alloy under varying process parameters, producing both defect-free and defective parts. These parts were classified based on surface appearance and welding camera images. We collected welding current and voltage data at a 5 kHz sampling rate and extracted features from both time and frequency domains using a knowledge-based approach. Isolation Forest, k-Nearest Neighbor, Artificial Neural Network, XGBoost, and LGBM models were trained on these features, and the results shown best performance of boosting models, achieving F1 scores of 0.927 and 0.945, respectively. These models presented higher performance compared to other models like k-Nearest Neighbor, whereas Isolation Forest and Artificial Neural Network posses lower performance due to overfitting, with an F1 score of 0.507 and 0.56, respectively. Then, by leveraging the feature importance capabilities of these models, we identified key signal characteristics that distinguish between normal and anomalous behavior, improving the explainability of the detection process and in general about the process physics. [ABSTRACT FROM AUTHOR]

Background/Objectives: Predicting patient readmission is an important task for healthcare risk management, as it can help prevent adverse events, reduce costs, and improve patient outcomes. In this paper, we compare various conventional machine learning models and deep learning models on a multimodal dataset of electronic discharge records from an Irish acute hospital. Methods: We evaluate the effectiveness of several widely used machine learning models that leverage patient demographics, historical hospitalization records, and clinical diagnosis codes to forecast future clinical risks. Our work focuses on addressing two key challenges in the medical fields, data imbalance and the variety of data types, in order to boost the performance of machine learning algorithms. Furthermore, we also employ SHapley Additive Explanations (SHAP) value visualization to interpret the model predictions and identify both the key data features and disease codes associated with readmission risks, identifying a specific set of diagnosis codes that are significant predictors of readmission within 30 days. Results: Through extensive benchmarking and the application of a variety of feature engineering techniques, we successfully improved the area under the curve (AUROC) score from 0.628 to 0.7 across our models on the test dataset. We also revealed that specific diagnoses, including cancer, COPD, and certain social factors, are significant predictors of 30-day readmission risk. Conversely, bacterial carrier status appeared to have minimal impact due to lower case frequencies. Conclusions: Our study demonstrates how we effectively utilize routinely collected hospital data to forecast patient readmission through the use of conventional machine learning while applying explainable AI techniques to explore the correlation between data features and patient readmission rate. [ABSTRACT FROM AUTHOR]

Background: Depression is a pervasive mental health condition, particularly affecting older adults, where early detection and intervention are essential to mitigate its impact. This study presents an explainable multi-layer dynamic ensemble framework designed to detect depression and assess its severity, aiming to improve diagnostic precision and provide insights into contributing health factors. Methods: Using data from the National Social Life, Health, and Aging Project (NSHAP), this framework combines classical machine learning models, static ensemble methods, and dynamic ensemble selection (DES) approaches across two stages: detection and severity prediction. The depression detection stage classifies individuals as normal or depressed, while the severity prediction stage further classifies depressed cases as mild or moderate-severe. Finally, a confirmation depression scale prediction model estimates depression severity scores to support the two stages. Explainable AI (XAI) techniques are applied to improve model interpretability, making the framework more suitable for clinical applications. Results: The framework's FIRE-KNOP DES algorithm demonstrated high efficacy, achieving 88.33% accuracy in depression detection and 83.68% in severity prediction. XAI analysis identified mental and non-mental health indicators as significant factors in the framework's performance, emphasizing the value of these features for accurate depression assessment. Conclusions: This study emphasizes the potential of dynamic ensemble learning in mental health assessments, particularly in detecting and evaluating depression severity. The findings provide a strong foundation for future use of dynamic ensemble frameworks in mental health assessments, demonstrating their potential for practical clinical applications. [ABSTRACT FROM AUTHOR]

Simple Summary: A novel mammogram image-based BCa explainable AI (BCaXAI) model is proposed, with a deep learning-based framework, for precise, noninvasive, objective, and timely manner diagnosis of BCa. The proposed BCaXAI leverages the Inception-ResNet v2 architecture, where the integration of explainable AI components, such as Grad-CAM, provides radiologists with valuable visual insights into the model's decision-making process, fostering trust and confidence in the AI-based system. Breast cancer (BCa) poses a severe threat to women's health worldwide as it is the most frequently diagnosed type of cancer and the primary cause of death for female patients. The biopsy procedure remains the gold standard for accurate and effective diagnosis of BCa. However, its adverse effects, such as invasiveness, bleeding, infection, and reporting time, keep this procedure as a last resort for diagnosis. A mammogram is considered the routine noninvasive imaging-based procedure for diagnosing BCa, mitigating the need for biopsies; however, it might be prone to subjectivity depending on the radiologist's experience. Therefore, we propose a novel, mammogram image-based BCa explainable AI (BCaXAI) model with a deep learning-based framework for precise, noninvasive, objective, and timely manner diagnosis of BCa. The proposed BCaXAI leverages the Inception-ResNet V2 architecture, where the integration of explainable AI components, such as Grad-CAM, provides radiologists with valuable visual insights into the model's decision-making process, fostering trust and confidence in the AI-based system. Based on using the DDSM and CBIS-DDSM mammogram datasets, BCaXAI achieved exceptional performance, surpassing traditional models such as ResNet50 and VGG16. The model demonstrated superior accuracy (98.53%), recall (98.53%), precision (98.40%), F1-score (98.43%), and AUROC (0.9933), highlighting its effectiveness in distinguishing between benign and malignant cases. These promising results could alleviate the diagnostic subjectivity that might arise as a result of the experience-variability between different radiologists, as well as minimize the need for repetitive biopsy procedures. [ABSTRACT FROM AUTHOR]

With the growing development and deployment of automated vehicles (AVs), it is crucial to understand the associated risks and factors contributing to collisions involving AVs. California possesses an immense amount of publicly available data from AV testing due to the requirement laid out by the California Department of Motor Vehicles (DMV), which requires all automated vehicle operators to report collisions for any level of severity. However, this information is reported in specific forms and requires a laborious task to aggregate data from these reports. This study creates an automated data extraction system for these reports and analyses collision characteristics using logistic regression models as well as XGBoost models with SHapley Additive exPlanation (SHAP) interpretation. Additionally, these characteristics are matched with the characteristics of non-automated vehicles (non-AVs) for the same region. The results of the study indicate that rear-end collisions are the most common type of collision observed in currently deployed AVs. The analysis further revealed an increased likelihood of injury-prone rear-end collisions in AVs at intersections compared to non-AVs. Transportation policymakers and researchers should take these safety concerns into account when addressing AV deployment and developing appropriate measures to mitigate collision risks in mixed fleet conditions. [ABSTRACT FROM AUTHOR]

The common black box nature of machine learning models is an obstacle to their application in health care context. Their widespread application is limited by a significant "lack of trust." So, the main goal of this work is the development of an evaluation approach that can assess, simultaneously, trust and performance. Trust assessment is based on (i) model robustness (stability assessment), (ii) confidence (95% CI of geometric mean), and (iii) interpretability (comparison of respective features ranking with clinical evidence). Performance is assessed through geometric mean. For validation, in patients' stratification in cardiovascular risk assessment, a Portuguese dataset (N=1544) was applied. Five different models were compared: (i) GRACE score, the most common risk assessment tool in Portugal for patients with acute coronary syndrome; (ii) logistic regression; (iii) Naïve Bayes; (iv) decision trees; and (v) rule-based approach, previously developed by this team. The obtained results confirm that the simultaneous assessment of trust and performance can be successfully implemented. The rule-based approach seems to have potential for clinical application. It provides a high level of trust in the respective operation while outperformed the GRACE model's performance, enhancing the required physicians' acceptance. This may increase the possibility to effectively aid the clinical decision. [ABSTRACT FROM AUTHOR]

Lead generation refers to the identification of potential topics (the 'leads') of importance for journalists to report on. In this article we present μ XL, a new lead generation tool based on a microservice architecture that includes a component of explainable AI. μ XL collects and stores historical and real-time data from web sources, like Google Trends, and generates current and future leads. Leads are produced by a novel engine for hypothetical reasoning based on temporal logical rules, which can identify propositions that may hold depending on the outcomes of future events. This engine also supports additional features that are relevant for lead generation, such as user-defined predicates (allowing useful custom atomic propositions to be defined as Java functions) and negation (needed to specify and reason about leads characterized by the absence of specific properties). Our microservice architecture is designed using state-of-the-art methods and tools for API design and implementation, namely API patterns and the Jolie programming language. Thus, our development provides an additional validation of their usefulness in a new application domain (journalism). We also carry out an empirical evaluation of our tool. [ABSTRACT FROM AUTHOR]

Deep neural networks are increasingly employed to model classroom dialogue and provide teachers with prompt and valuable feedback on their teaching practices. However, these deep learning models often have intricate structures with numerous unknown parameters, functioning as black boxes. The lack of clear explanations regarding their classroom dialogue analysis likely leads teachers to distrust and underutilize these AI‐powered models. To tackle this issue, we leveraged explainable AI to unravel classroom dialogue analysis and conducted an experiment to evaluate the effects of explanations. Fifty‐nine pre‐service teachers were recruited and randomly assigned to either a treatment (n = 30) or control (n = 29) group. Initially, both groups learned to analyse classroom dialogue using AI‐powered models without explanations. Subsequently, the treatment group received both AI analysis and explanations, while the control group continued to receive only AI predictions. The results demonstrated that teachers in the treatment group exhibited significantly higher levels of trust in and technology acceptance of AI‐powered models for classroom dialogue analysis compared to those in the control group. Notably, there were no significant differences in cognitive load between the two groups. Furthermore, teachers in the treatment group expressed high satisfaction with the explanations. During interviews, they also elucidated how the explanations changed their perceptions of model features and attitudes towards the models. This study is among the pioneering works to propose and validate the use of explainable AI to address interpretability challenges within deep learning‐based models in the context of classroom dialogue analysis.Practitioner notesWhat is already known about this topicClassroom dialogue is recognized as a crucial element in the teaching and learning process.Researchers have increasingly utilized AI techniques, particularly deep learning methods, to analyse classroom dialogue.Deep learning‐based models, characterized by their intricate structures, often function as black boxes, lacking the ability to provide transparent explanations regarding their analysis. This limitation can result in teachers harbouring distrust and underutilizing these models.What this paper addsThis paper highlights the importance of incorporating explainable AI approaches to tackle the interpretability issues associated with deep learning‐based models utilized for classroom dialogue analysis.Through an experimental study, this paper demonstrates that providing model explanations enhances teachers' trust in and technology acceptance of AI‐powered classroom dialogue models, without increasing their cognitive load.Teachers express satisfaction with the model explanations provided by explainable AI.Implications for practice and/or policyThe integration of explainable AI can effectively address the challenge of interpretability in complex AI‐powered models used for analysing classroom dialogue.Intelligent teaching systems designed for classroom dialogue can benefit from advanced AI models and explainable AI approaches, which offer users both automated analysis and clear explanations. By enabling users to understand the underlying rationale behind the analysis, the explanations can contribute to fostering trust and acceptance of the AI models among users. [ABSTRACT FROM AUTHOR]

In many regions globally, snowmelt‐recharged mountainous karst aquifers serve as crucial sources for municipal and agricultural water supplies. In these watersheds, complex interplay of meteorological, topographical, and hydrogeological factors leads to intricate recharge‐discharge pathways. This study introduces a spatially distributed deep learning precipitation‐runoff model that combines Convolutional Long Short‐Term Memory (ConvLSTM) with a spatial attention mechanism. The effectiveness of the deep learning model was evaluated using data from the Logan River watershed and subwatersheds, a characteristically karst‐dominated hydrological system in northern Utah. Compared to the ConvLSTM baseline, the inclusion of a spatial attention mechanism improved performance for simulating discharge at the watershed outlet. Analysis of attention weights in the trained model unveiled distinct areas contributing the most to discharge under snowmelt and recession conditions. Furthermore, fine‐tuning the model at subwatershed scales provided insights into cross‐subwatershed subsurface connectivity. These findings align with results obtained from detailed hydrogeochemical tracer studies. Results highlight the potential of the proposed deep learning approach to unravel the complexities of karst aquifer systems, offering valuable insights for water resource management under future climate conditions. Furthermore, results suggest that the proposed explainable, spatially distributed, deep learning approach to hydrologic modeling holds promise for non‐karstic watersheds. Key Points: An explainable spatially distributed deep learning hydrologic model is built using a spatial attention mechanismThe model trained with streamflow at the watershed outlet simulates discharge at subwatershed scales reasonably wellRecharge‐discharge pathways suggested by attention weights are mostly consistent with hydrogeochemical tracer studies [ABSTRACT FROM AUTHOR]