Search Results (197)

Systems that can reliably detect intrusion are increasingly in demand as the scale and heterogeneity of Internet of Things (IoT) networks rise. However, when strong tabular models are employed as a benchmark, it is not clear whether the complexity of ensembling actually pays off. The aim of this work is to systematically assess the reliability and accuracy of soft-voting ensembles across three benchmark datasets (CICIoT2023, TON_IoT, and Edge-IIoTset), applying a leakage-free protocol with twice-repeated stratified 10-fold cross-validation. Ensembles of varying sizes and compositions were benchmarked against Decision Tree, KNN, Random Forest, LightGBM, XGBoost, and CatBoost and the results revealed that ensemble complexity did not lead to a consistent increase in performance. For example, SoftVote-2 increased macro-F1 over LightGBM on CICIoT2023 from 0.8506 to 0.8563, whereas LightGBM remained superior on TON_IoT and Edge-IIoTset. A leakage analysis demonstrated that resampling before data partitioning increased accuracy by around 8 percentage points and macro-F1 by 14 to 17 percentage points. To assess the trade-off between performance and complexity, training time, inference latency, memory, and model size were all evaluated, and LIME and SHAP were also assessed for explanation stability, local fidelity, and attribution agreement. XAI-guided feature selection showed that the 15 most important features retained 98.4–99.4% of the original macro-F1 and decreased inference latency by up to 38%. The results indicate that the value of ensemble complexity varies by dataset and should be weighed against its computational cost. The main contribution of this study is a leakage-aware and explainability-informed framework that can be used to judge when ensemble complexity yields genuine predictive and practical benefits for the detection of IoT intrusion. Full article

Accurate maps of perennial crop types support agricultural monitoring, water-use accounting, and strategic decisions on climate adaptation. Hyperspectral imagery offers the fine spectral detail needed to separate spectrally similar plant species, but its high dimensionality complicates classification. This study assessed the discrimination of spectrally similar perennial crop types using AVIRIS-NG imagery and random forest classification. Three perennial crop type classification schemes (with 17, nine, and six classes, respectively) were targeted. Classification models were trained on labelled crop type samples collected in the Western Cape Province of South Africa. Feature selection (recursive feature elimination and Boruta) and feature extraction (principal component analysis and the minimum noise fraction transform) were used to reduce image dimensionality. Feature extraction improved accuracy over the spectral baseline across all classification schemes, with the minimum noise fraction transform consistently outperforming principal component analysis, including in the 17-class scheme, and feature extraction alone yielding the highest accuracies. Overall accuracy increased with increased aggregation, from approximately 66% in the 17-class scheme to 75% in the six-class scheme. Spectrally distinct crops such as cherry and macadamia were classified reliably, whereas structurally similar tree crops (e.g., lemon and lime, peach and plum) remained poorly classified. The findings suggest that hyperspectral data holds much potential for differentiating between spectrally similar perennial crop types but that classification schemes must be carefully designed to reduce misclassifications. Full article

Deep visual recognition has achieved remarkable success across various domains, including medical imaging, autonomous vehicles, and security systems. However, the black-box nature of deep learning models poses challenges in terms of transparency and trust, especially in critical applications where human understanding is essential. Explainable AI (XAI) seeks to address these concerns by providing human-interpretable explanations for model predictions. This review explores the key techniques for explainability in deep visual recognition, including model-agnostic methods such as LIME and SHAP, model-specific approaches like saliency maps and feature visualization, and intrinsically interpretable models like decision trees and rule-based systems. We also discuss the evaluation of explainability through metrics like fidelity, consistency, and stability, and explore the challenges of balancing model performance with interpretability. Furthermore, we examine applications of XAI in medical imaging, autonomous driving, security and surveillance, agriculture, satellite imagery and remote sensing, industrial inspection, and visual forensics, highlighting how domain-specific data and operational constraints affect the required form and validation of explanations. Finally, we address current research gaps and propose future directions for enhancing the robustness and human–AI interaction in explainable visual recognition systems. As AI continues to be integrated into safety-critical domains, the development of explainable, transparent, and trustworthy AI systems will be crucial for their widespread adoption and ethical use. Full article

Machine learning has become an important tool in plant genomic prediction for modeling complex genotype–phenotype relationships and improving breeding decisions. However, many high-performing models, particularly ensemble and deep learning approaches, remain difficult to interpret, limiting their biological applicability. This review summarizes major machine learning methods and explainable artificial intelligence (XAI) approaches used in plant genomics, including SHAP (SHapley Additive exPlanations), LIME (Local Interpretable Model-Agnostic Explanations), attention mechanisms, permutation importance, tree-based feature importance, and gradient-based attribution methods. XAI can help identify influential SNPs, genomic regions, candidate genes, regulatory elements, and omics features associated with complex traits. For example, SHAP analysis in an almond germplasm collection identified a genomic region associated with shelling fraction, illustrating how XAI can generate testable hypotheses for further validation. The review further discusses applications in trait prediction, breeding, functional genomics, and multi-omics integration. Importantly, we emphasize major limitations, including data bias, model instability, correlated genomic markers, limited model transferability, and the common misconception that feature importance implies biological causality. We recommend integrating XAI with linkage disequilibrium pruning, stability assessment, biological annotation, and experimental validation before prioritizing candidate genes. Overall, XAI should be considered a framework for model interpretation, feature prioritization, and hypothesis generation rather than a replacement for experimental validation in plant genomics and breeding. Full article

Background: Hepatitis C virus (HCV) affects approximately 50 million people worldwide and progresses silently through distinct hepatic stages, yet most machine learning staging approaches offer no clinician-facing explanation and rarely evaluate whether explanations hold across architectures. Methods: We trained five models—Logistic Regression, Random Forest, XGBoost, LightGBM, and SVM—on the UCI HCV dataset (615 patients, four classes after merging the seven-instance suspect blood-donor group into a single Donor/Control class) and applied SHAP, LIME, and DiCE counterfactuals. A cross-model Spearman agreement analysis quantified feature ranking consistency, with a leakage-controlled pipeline and five-fold stratified cross-validation applied. Results: In the main 5-fold cross-validation, LightGBM achieved the highest macro-F1 (0.684 ± 0.031). Under repeated-seed cross-validation, XGBoost and LightGBM gave closely matched values (0.648 ± 0.023 and 0.645 ± 0.026), indicating comparable robustness among the boosted-tree models, both marginally ahead of the remaining three, with Random Forest close behind. Logistic Regression reached the highest macro-F1 on the single held-out split (0.790), but its cross-validated score was markedly lower (0.598 ± 0.063), underlining how unstable single-split estimates are in a small, imbalanced cohort. SHAP consistently identified AST, GGT, CHE, and ALP across the tree-based models, with CHE emerging as the leading Cirrhosis-stage marker in the boosted models. Cross-model Spearman correlations reached 0.923–0.958 among tree-based models; SHAP-LIME overlap ranged from 1/5 to 4/5. Conclusions: The framework identifies stage-specific biochemical importance patterns consistent with known HCV disease progression. The convergent finding on CHE for Cirrhosis, supported by SHAP in the tree-based models and appearing among the top LIME features in three of the five model-specific Cirrhosis explanations, supports CHE as a candidate marker worth evaluating in advanced disease. Full article

Background: Dengue fever remains a major public health challenge in endemic regions, where recovery duration varies considerably across patients due to a combination of clinical, demographic, and contextual factors. Although machine learning (ML) approaches have increasingly been applied to dengue related prediction tasks, many existing models operate as black boxes, limiting their interpretability and practical usefulness in healthcare settings. This study presents an Explainable Artificial Intelligence (XAI) based machine learning framework for analyzing dengue recovery duration using a multi-center clinical dataset collected from healthcare institutions across Khyber Pakhtunkhwa, Pakistan. Methods: Clinical records from 100 laboratory-confirmed dengue patients treated across multiple healthcare institutions were analyzed. The dataset included demographic, socio-economic, and clinical variables. Four machine learning models: Linear Regression, Decision Tree, Random Forest, and Neural Network, were developed and evaluated using 10-fold cross-validation. Explainability techniques, including Partial Dependence Plots (PDP), Individual Conditional Expectation (ICE), and Local Interpretable Model-Agnostic Explanations (LIME), were employed to investigate global and patient specific factors influencing recovery duration. Results: Among the evaluated models, Random Forest demonstrated the best overall predictive performance, achieving the lowest Root Mean Square Error (RMSE; 11.29 days) and Mean Absolute Error (MAE; 9.09 days), corresponding to a 40.4% reduction in prediction error compared with Linear Regression. Decision Tree also showed substantial improvement, reducing RMSE by 37%, whereas the Neural Network achieved a more modest improvement of 8.6%. Although all models exhibited relatively low coefficient of determination (R²) values (maximum R² = 0.026), the explainability analyses consistently identified age and platelet count as the most influential predictors of recovery duration. Older age and lower platelet counts were generally associated with longer recovery periods, while hospital type, education level, and blood group also contributed to prediction outcomes. ICE and LIME analyses further revealed considerable patient level heterogeneity, indicating that recovery trajectories are shaped by complex interactions among clinical, demographic, and contextual factors rather than a single dominant predictor. Full article

Background/Objectives: Chronic kidney disease (CKD) is an incurable and progressive condition; if diagnosed at an early stage, it would significantly reduce the risk of complications and enhance the outcomes for the patient. Methods: In this study, a custom dataset of 380 instances and 20 clinical attributes was used to develop and evaluate the machine learning (ML) models for reliable CKD prediction and to enhance the interpretability using explainable artificial intelligence (XAI) techniques. Artificial neural networks, C5.0, CHAID, logistic regression, linear support vector machines (L1 and L2 regularization), k-nearest neighbors (KNN), random tree, and deep neural networks were implemented. Correlation-based methods, recursive feature elimination, and LASSO were used for feature selection. SMOTE and SMOTETomek resampling techniques were used to address class imbalance. Three experimental set-ups were considered: (i) using SMOTETomek, (ii) with and without SMOTE, and (iii) grouped features according to the strength of correlation (high, moderate, low). Accuracy, precision, recall, F1 Score, AUC, and Gini index were used to evaluate the model’s performance. The pipeline was implemented in Python using the scikit-learn and imbalanced-learn packages. Results: Using SHAP and LIME, model interpretability was improved, with the KNN classifier obtaining the highest accuracy of 94.74% without SMOTE, and the C5.0 model obtained the highest accuracy of 92.98% with SMOTE. In the feature-group experiments, the L1-regularized linear SVM achieved high accuracy (89.47%) with highly correlated features. In general, both resampling methods improved model robustness, and feature selection methods reduced the model’s dimensionality with little loss in performance. Conclusions: The ML framework proposed is promising in predicting CKD with high accuracy and interpretability with relevance. By combining feature selection with class balancing and explainable AI, the model’s performance improves, and its clinical trustworthiness is enhanced. The results indicate the potential in using ML-based decision support systems for early-stage CKD diagnosis and personalized healthcare. Full article

The growing worldwide prevalence of Diabetes Mellitus highlights the urgent need for effective early detection methods to enable prompt intervention. This study develops a machine learning-based decision-support prototype for predicting diabetes risk using health metrics from the DiaHealth dataset, a recently published Bangladeshi open-source dataset for Type 2 diabetes prediction. Five supervised learning algorithms were evaluated: Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbour (KNN), Decision Tree (DT), and Random Forest (RF). Models were assessed across three stages: before feature scaling, after standardisation, and following hyperparameter optimisation via GridSearchCV, using accuracy, precision, recall, and F1-score as evaluation metrics. LR and SVM showed marked improvements after standardisation, consistent with their sensitivity to feature magnitude, whilst tree-based approaches such as DT and RF remained largely unchanged. KNN displayed minimal sensitivity to scaling, which is discussed in relation to the feature distributions of the dataset. Following hyperparameter tuning, RF achieved the highest accuracy of 95%, outperforming all other models. RF predictions were interpreted using Local Interpretable Model-agnostic Explanations (LIME) to promote transparency in model decision-making. The best-performing model was subsequently deployed as an interactive web-based prototype application using Streamlit, providing real-time prediction outputs. These findings demonstrate how preprocessing choices and hyperparameter tuning can differentially affect algorithm performance and illustrate the potential of combining explainable AI with practical deployment for diabetes risk assessment in a research context. Full article

Thirteen mandarin and hybrid cultivars grafted onto the commercial rootstocks Rangpur Lime and Swingle Citrumelo were comparatively assessed for vegetative growth, fruit physicochemical attributes, and field incidence and severity of Altenaria Brown Spot (ABS) and Huanglongbing (HLB). The experiment was conducted from January 2015 to December 2018 under a randomized block design, with ten replicates per scion–rootstock combination. Plant height, canopy volume, fruit mass, juice yield, acidity, soluble solids, and disease assessments were performed. RL induced greater vegetative growth but was associated with higher HLB severity, particularly in the Dekopon IAC 2009 and TM × LP 358 varieties. SC resulted in less vigorous trees but improved fruit quality, with higher acidity and soluble solids. Regarding ABS, the Loose Jacket IAC 515 and Muscia varieties showed high susceptibility, while Ortanique IAC 554 and Rainha BRS exhibited tolerance to both ABS and HLB. These findings suggest that although RL promotes vigorous growth, it may increase disease susceptibility, whereas SC is associated with reduced disease severity and improved fruit quality. Ortanique IAC 554 and Rainha BRS showed consistently low severity of ABS and HLB, combined with stable vegetative development and fruit quality, underscoring the importance of rootstock choice for guiding cultivar deployment in orchards under high disease pressure. Full article

Colorectal cancer (CRC) remains one of the leading causes of cancer-related mortality worldwide, highlighting the critical need for non-invasive, accurate, and interpretable diagnostic tools. Metabolomic profiling of fecal microbial extracellular vesicles (EVs) offers a promising yet underexplored avenue for biomarker discovery when integrated with explainable machine learning (ML) frameworks. This study aimed to identify stool-derived microbial EV metabolite biomarkers that discriminate CRC patients from healthy controls and to develop interpretable ML classifiers for non-invasive CRC detection. Metabolomic profiles of fecal microbial EVs from 76 age- and sex-comparable participants (36 CRC, 40 controls) were obtained using LC/QTOFMS and GC/TOFMS. Three ML classifiers (TPOT, LightGBM, XGBoost) were trained and evaluated through 100-repeat stratified hold-out and nested 5-fold cross-validation, with SHAP and LIME applied for global and local interpretability. Fourteen metabolites were significantly dysregulated between the CRC and control groups (adjusted p < 0.05), with 13 upregulated and one (aminoisobutyric acid) downregulated. Furoic acid exhibited perfect diagnostic discrimination, followed by palmitic acid and tyramine. Nested cross-validation demonstrated robust performance: TPOT achieved AUC = 0.997 ± 0.005, sensitivity = 0.973 ± 0.022, and MCC = 0.957 ± 0.033. Hold-out validation corroborated these findings (AUC = 0.998 ± 0.008). SHAP analysis identified furoic acid, palmitic acid, and tyramine as the dominant predictive features, while aminoisobutyric acid exhibited a distinctive protective pattern. LIME analysis corroborated these findings at the individual prediction level. The identified fecal EV-derived metabolite panel—particularly furoic acid, palmitic acid, and tyramine—shows strong potential to predict CRC in a non-invasive, interpretable manner; however, given the modest sample size, these findings should be considered hypothesis-generating and require validation in larger, prospective, multi-center cohorts before clinical translation. Full article

This study investigates the effect of nano-zeolite and lime on the resistance of reconstituted soil using an integrated experimental and explainable machine learning framework. Soil samples were prepared with varying proportions of nano-zeolite, lime, and fines, and cured under controlled temperature and time conditions. Soil resistance (q) was measured to evaluate the mechanical performance of each mixture. Eight machine learning models, including artificial neural networks (ANN), random forest (RF), random tree (RT), random committee–random tree (RC-RT), M5Rules, KStar, RBFS, and additive regression–decision stump (AR-DS), were developed using Weka 3.8.6 to predict soil resistance based on the input parameters. Model performance was assessed using SSE, MAE, MSE, RMSE, Error %, Accuracy %, R², correlation coefficient, Willmott Index, Nash–Sutcliffe Efficiency, Kling–Gupta Efficiency, and SMAPE. ANN and RF achieved superior accuracy (R² ≥ 0.98) with minimal prediction error, effectively capturing the nonlinear interactions between stabilizer content, curing time, and environmental conditions. Sensitivity analyses using the analysis index and SHAP values revealed that nano-zeolite, lime, and curing time were the dominant factors influencing soil resistance, while fines content and curing temperature had secondary effects. The results demonstrate that nano-zeolite and lime significantly enhance soil resistance and that explainable machine learning models can reliably predict and interpret soil performance, providing a data-driven framework for optimized soil stabilization in geotechnical engineering applications. Full article

Existing post hoc XAI (Explainable Artificial Intelligence) methods produce numerical attributions without symbolic structure (SHAP, LIME), low-coverage local rules (Anchors), or crisp tree surrogates without an interpretable rule-level uncertainty proxy. We present a fuzzy rule-based explanation framework for tabular black-box classifiers, extracting global IF–THEN rules with linguistic labels. This was validated on a 13-dataset benchmark with four model families (Wilcoxon, Friedman, TOST equivalence): (i) prediction-boundary-aware fuzzy partitioning raises mean fidelity from a vanilla Wang–Mendel baseline of 0.736 to 0.893 (+10.4 pp excluding the Breast Cancer outlier; +15.7 pp aggregate, both transparently reported); (ii) fired-rule consequent entropy provides a zero-cost rule-level uncertainty proxy (Spearman ρ = 0.420 with model prediction entropy, significant on 11/12 datasets—moderate by Cohen’s convention, with a 4/12 weak-correlation tail; complementary to probability-entropy and margin baselines). Fidelity is statistically equivalent to tree surrogates on classification (TOST p = 0.002, δ = 0.05) at ≈100% coverage. SHAP/LIME are excluded from the formal stability ranking because the perturbation metric measures the wrapped black-box rather than the attribution vector; cross-explainer comparison is reported in grouped form (full-coverage surrogates vs. local-coverage methods). On continuous regression (California Housing fidelity 0.422 vs. TreeSurrogate 0.840) and XOR-type multi-feature interactions, the framework is structurally weaker, addressed by a planned TSK extension. Full article

Urban air pollution results from complex interactions between vehicle emissions, meteorological conditions, and atmospheric chemistry. While machine learning models achieve high accuracy in air quality prediction, their limited transparency hinders policy adoption. We present an integrated (M-ETAQI) framework combining multiple XAI techniques, temporal decomposition, and causal inference to quantify traffic and meteorological contributions to PM₁₀, PM_2.5, NO_X, and NO₂ concentrations in the Istanbul FSM Bridge corridor (2022–2023 hourly data). Five machine learning models, including XGBoost, LightGBM, CatBoost, Random Forest, and CNN–LSTM–Attention, were trained with temporal cross-validation. SHAP, LIME, PDP, and ALE were applied for interpretability; STL decomposition isolated temporal components, and CCM tested causal links. Tree-based models achieved R² > 0.80 for all pollutants, with CatBoost reaching PM_2.5 R² = 0.876. SHAP confirmed Lag1 as the dominant feature. Wind speed had a significant negative effect on NO_X, while traffic contributed ~20% to NO_X, twice that of other pollutants. STL showed the trend component dominated total variance; NO₂ trend variance = 56.3%. CCM revealed wind speed as the strongest causal driver of NO_X (ρ = 0.37) and confirmed direct traffic–NO_X links. Knowledge distillation from CatBoost improved CNN–LSTM–Attention performance. The four XAI methods yielded consistent attributions, providing robust, cross-validated evidence for traffic management and air-quality policy. Full article

Urinary Tract Infections (UTIs) represent one of the most prevalent bacterial infections globally, posing significant health burdens, especially in low- and middle-income countries (LMICs), due to delayed diagnoses, limited access to laboratory services, and rising antimicrobial resistance. This study presents a machine learning (ML)-based diagnostic support framework for early UTI detection, leveraging structured clinical data and explainable artificial intelligence (XAI) techniques to enhance interpretability and trust among healthcare providers. A patient dataset containing 4865 records was used in the study to train and test Extreme Gradient Boosting (XGBoost), Decision Tree (DT) and Random Forest (RF) classifiers, while class imbalance was addressed using Synthetic Minority Over-sampling Technique (SMOTE). The performance of the models was evaluated through accuracy, precision, recall, F1-score, Log Loss, and AUC-ROC, and random forest showed the best results (accuracy: 86.43%, F1-score: 86.71%, AUC-ROC: 0.8695). To ensure that such models can be adopted by stakeholders in the health sector, Local Interpret-able Model-agnostic Explanations (LIME) were integrated, which identified painful urination, urinary frequency, and suprapubic pain as primary predictors in the model. This study shows that interpretable ML models can be helpful in resource-limited regions in predicting UTIs, thereby rendering a solution to improve the management of infections in these regions. Full article

Rootstock strongly influences citrus tree performance, but information remains limited for some regionally important cultivars. ‘IAPAR 73’, an early-season sweet orange commonly grown in Paraná, Brazil, has not been previously evaluated for rootstock responses. This study assessed the long-term effects of nine rootstocks, including ‘Rangpur’ lime, ‘Swingle’ citrumelo, ‘Volkamer’ lemon, ‘Caipira DAC’ and ‘Trifoliate’ oranges, ‘Cleopatra’ and ‘Sunki’ mandarins, ‘Carrizo’ and ‘Fepagro C-13’ citranges, on vegetative growth, yield, production stability, and fruit quality under Brazilian subtropical conditions. Tree growth was monitored annually for 10 years (2003–2013) and analyzed at establishment (5 years) and full production (10 years) phases of the orchard. Yield and fruit quality were evaluated across multiple harvests, and total soluble solids (TSS) stability was quantified using the coefficient of variation. Rootstock effects were analyzed using linear mixed-effects models in a randomized complete block design, considering rootstock and year as fixed effects and blocks as random effects. Rootstock significantly influenced all evaluated traits. ‘Carrizo’, ‘Cleopatra’, ‘Sunki’, and ‘Caipira DAC’ induced vigorous canopy growth and higher cumulative yields to the scion, while ‘Volkamer’ showed high yield efficiency and production stability. ‘Swingle’ and ‘Trifoliate’ enhanced TSS, TSS/TA ratios, and juice quality stability but induced lower vigor and yield, similar to ‘Rangpur’. This study provides the first evidence-based guidance for ‘IAPAR 73’ production, demonstrating that rootstock diversification can maximize productivity, stability, and sustainability in citrus orchards. Full article