To compare machine learning models using different combinations of clinical and imaging variables for classifying ischemic stroke patients as having an onset-to-imaging (OTI) time within or beyond 4.5 h. We analyzed 993 patients with known OTI time from the MR CLEAN Registry and LATE trial. Data were split into training and test sets (80:20). We developed models using various combinations of variables to classify OTI time, including clinical-radiological information, and variables automatically extracted from segmented ischemic regions on non-contrast CT, such as net water uptake (NWU), lesion volume, and radiomics features. Performance was assessed using the area under the receiver operating characteristic curve (AUC). Of 993 patients, 199 (20%) presented beyond 4.5 h. The model including only clinical-radiological scores, and the one including only NWU, achieved an AUC of 0.65. Performance was higher for models that included NWU combined with lesion volume or clinical-radiological scores (AUCs ranging from 0.70 to 0.75). Radiomics-based models achieved the highest performance with AUCs of 0.81, significantly outperforming NWU-based models. Key predictors for identifying patients beyond 4.5 h included homogeneous lesion textures in both core and hypoperfused areas, smaller hypoperfused area volumes, higher core NWU, and lower baseline NIHSS scores. We found that radiomics-based models outperform models including NWU measurements for classifying stroke OTI time in this endovascular therapy population. The superior performance suggests that texture, shape, and intensity patterns of ischemic lesions may capture more information about lesion age than single metrics like NWU. External validation in broader stroke populations is needed to establish clinical utility. Question Which combinations of clinical and CT-derived variables enable the most accurate classification of stroke onset time within versus beyond 4.5 h using machine learning? Findings Models using radiomics features achieved superior accuracy (AUC 0.81) compared to models using net water uptake measurements (AUC 0.65) for onset time classification. Clinical relevance Automated CT-based radiomics models enable accurate stroke onset time classification without advanced imaging, potentially expanding treatment options for patients with unknown symptom onset times in centers lacking MRI capabilities.

Machine learning model training, which ultimately optimizes a model’s cost function is usually a time- consuming and computationally intensive process on classical computers. This has been more intense due to the in- creased demand for large-scale data analysis, requiring unconventional computing paradigms like quantum computing to enhance training efficiency. Adiabatic quantum computers have excelled at solving optimization problems, which require the quadratic unconstrained binary optimization (QUBO) format of the problem of interest. In this study, the squared error minimization in the multiple linear regression model is reformulated as a QUBO problem enabling it to be solved using D-wave adiabatic quantum computers. Same formulation was used to obtain a solution using gate-based algorithms such as quantum approximate optimization algorithm (QAOA) and sampling variational quantum eigensolver (VQE) im- plemented via IBM Qiskit. The results obtained through these approaches in the context of runtime and mean squared error(MSE) were analyzed and compared to the classical approaches. Our experimental results indicate a runtime ad- vantage in the D-wave annealing approach over the classical Scikit learn regression approach. The time advantage can be observed when N>524288 compared to Sklearn Linear Regression and when N>65536 compared to Sklearn SGDRegressor. Support vector machine induced neural networks, where the margin-based entropy loss is converted into a QUBO with Lagrangian approach is also focused in this study concerning the applicability for nonlinear models.

Introduction Parkinson’s disease (PD) is the second most common neurodegenerative disorder. The risk of frailty is significantly higher in patients with PD than in age-matched individuals without PD. This study aimed to develop a machine learning–based predictive model for frailty in PD. Methods We conducted a cross-sectional study of early- and middle-stage PD patients recruited from June 2024 to June 2025 at Shenzhen People’s Hospital. Frailty was assessed using the Fried criteria (five components: gait speed, grip strength, physical activity, fatigue, and weight loss). A total of 42 demographic and clinical variables, including disease history, Montreal cognitive assessment (MoCA), and unified Parkinson’s disease rating scale (MDS-UPDRS) scores, were collected and compared between PD patients with and without frailty. Spearman correlation and LASSO regression were used to identify independent risk factors. Multiple machine learning algorithms were applied to construct predictive models. Model performance was evaluated using receiver operating characteristic (ROC) curves, area under the ROC curve (AUC), decision curve analysis (DCA), calibration plots, and forest plots. Results A total of 205 PD patients were enrolled (133 non-frail, 72 frail; mean age non-frail 62.92 ± 9.69 years, frail 68.13 ± 8.44 years). Significant group differences were found in sex ( p = 0.013), age ( p < 0.001), disease severity (MDS-UPDRS, p < 0.001; modified Hoehn-Yahr stage (H&Y stage), p < 0.001), alcohol consumption ( p = 0.010), MoCA ( p < 0.001), HAMD ( p = 0.001), and Hamilton anxiety rating scale (HAMA) ( p < 0.001). Eight features were identified as independent predictors of frailty: sex, age, alcohol use, Modified H&Y stage, UPDRS-IV score, HAMA score, executive function, and naming. Among all tested algorithms, logistic regression achieved the best predictive performance (AUC = 0.83 in the test set), outperforming other machine learning models. Conclusion Frailty in PD was associated with female sex, older age, alcohol use, and more advanced disease severity. Patients with PD and frailty exhibited higher MDS-UPDRS scores, more severe cognitive impairment, and greater levels of depression and anxiety. Integrating clinical data with machine learning, especially logistic regression, provides a reliable and scalable tool for early identification and risk stratification of frailty in PD.

Abstract The rapid proliferation of AI-generated text from models such as ChatGPT-3.5 and ChatGPT-4 has raised critical challenges in verifying content authenticity and ensuring ethical use of language technologies. This study presents a comprehensive framework for distinguishing between human-written and GPT-generated text using a combination of machine learning, sequential deep learning, and transformer-based models. A balanced dataset of 20,000 samples was compiled, incorporating diverse linguistic and topical sources. Traditional algorithms and sequential architectures (LSTM, GRU, BiLSTM, BiGRU) were compared against advanced transformer models, including BERT, DistilBERT, ALBERT, and RoBERTa. Experimental findings revealed that RoBERTa achieved the highest performance (Accuracy = 96.1%), outperforming all baselines. Post-hoc temperature scaling (T = 1.476) improved calibration, while threshold tuning (t = 0.957) enhanced precision for high-stakes applications. McNemar’s test with Holm correction confirmed the statistical significance ( p < 0.05) of RoBERTa’s superiority. Efficiency analysis showed optimal trade-offs between accuracy and latency, and 20% pruning demonstrated sustainability potential. Furthermore, LIME and SHAP explainability analyses highlighted linguistic distinctions between AI-generated and human-authored text, and fine-grained error evaluation confirmed model robustness across text lengths. In conclusion, RoBERTa emerges as a reliable, interpretable, and computationally efficient model for detecting AI-generated content.

ABSTRACT Different fiber‐reinforcement structures in composites cause varied mechanical responses by affecting how stress is distributed throughout the sample under various loading conditions. Dry weft‐knitted fabric, characterized by its high stretch rate, exhibits different behaviors when used as a standalone fabric compared to when it is used as reinforcement in composites. The high‐rate elongation property of weft‐knitted fabric arises from the yarns' interaction and the loops' geometry within the fabric, not from the material used. This study examines the tensile behavior of dry weft‐knitted fabrics, intended to serve as reinforcement structures in composite materials. To analyze the impact of this structure, a mathematical model representing the 3D geometry of a Rib 1 × 1 weft‐knitted fabric was developed. The fabric geometry was modeled using Python scripting of the mathematical equations within the ABAQUS FE software for analyzing the modeled geometries. An experimental tensile test was conducted on a commercial Rib 1 × 1 weft‐knitted fabric in both the longitudinal and transverse directions. The fabric exhibited average elongations at break of 262% and 71%, and average maximum tensile forces of 132.8 and 395.6 N in the transverse and longitudinal directions, respectively. Remarkably, the mechanical response and deformation patterns of the numerical and experimental samples closely matched in these tests. Using mathematical models to recreate the complex geometries of these reinforcements in the dry fabric facilitated the accurate prediction of their mechanical responses using numerical analysis. These models are ideally suited for efficiently generating data to train machine learning and artificial intelligence models in the future.

Abstract Elastic constants govern how crystalline solids respond to stress, resist deformation, and transmit elastic waves. For cubic crystals, only three independent constants, , and , are required to connect microscopic elasticity with macroscopic mechanical behavior. These constants can be determined experimentally or through computational approaches such as molecular dynamics or density functional theory (DFT). However, experimental measurements are often challenging, while DFT calculations, though accurate, are computationally expensive. To address this, we have used a machine learning (ML) framework to predict and classify the elastic properties of ternary cubic alloys. A multi-output XGBoost model was trained on Matminer-featurized data from the Materials Project and validated against DFT calculations for 12 half-Heusler alloys, showing excellent agreement. We then applied the model to predict the elastic constants of 972 ternary materials, finding that 970 are mechanically stable and exhibit strong internal correlations among their elastic properties. Additionally, we employed unsupervised learning via K-means clustering to categorize these materials into four mechanical classes such as brittle and soft, ductile with moderate stiffness, stiff and strong, and highly ductile and soft. This integrated approach establishes a high-throughput pathway for accelerating the discovery and targeted design of materials with tailored mechanical performance.

Evaluating hybrid and machine learning models for landslide susceptibility mapping in a high-risk region of the Western Ghats, India

Introduction Endometrial cancer is the most common gynecological malignancy in high-income countries and lacks an established strategy for early detection. Prior studies suggest that the vaginal microbiome may hold diagnostic potential, but inconsistent findings have limited clinical translation. Methods We conducted a systematic review to collect and analyze vaginal 16S rRNA sequencing data from five independent cohorts (n = 265). These studies included women with histologically confirmed endometrial cancer and controls with benign gynecologic conditions. We used these datasets to identify microbial signatures associated with endometrial cancer and to develop a predictive machine learning model. Results Microbial diversity was significantly higher in endometrial cancer samples, and host characteristics influenced community composition. Peptoniphilus was reproducibly enriched in cancer samples across cohorts. An ensemble classifier accurately identified endometrial cancer in a held-out test set, achieving an area under the receiver operating characteristic curve of 0.93 (95% CI: 0.71–0.93), sensitivity of 1.0 (95% CI: 0.74–1.0), and a negative predictive value of 1.0 (95% CI: 0.59–1.0). Discussion These findings support the potential of vaginal microbiome profiling as a minimally invasive approach for early detection of endometrial cancer.

Objectives: To address the urgent need for forensic systems capable of detecting and analyzing advanced persistent threats in hybrid quantum-classical communication infrastructures, particularly those that may compromise quantum key distribution environments. Method: The study introduces a Quantum-Aware Forensics Investigation Framework, a multi-layered forensic architecture combining quantum telemetry, classical metadata analysis, and machine learning-driven threat classification. Experimental validation was conducted using a simulated testbed built with SimulaQron, Wireshark, and custom scripting tools. Various quantum attack scenarios were emulated, including intercept-resend, entanglement flooding, and control-plane hijacking. Machine learning models Random Forest, SVM, and Autoencoder were tested as standalone classifiers. A stacked ensemble model, with Random Forest and SVM as base learners and Logistic Regression as meta-classifier, was implemented for performance optimization. We used an experimentally generated, cross-layer dataset from a SimulaQron BB84 QKD emulation by combining quantum logs and classical control-plane captures under benign and scripted attacks such as intercept–resend, entanglement flooding, payload obfuscation, session hijacking, spoofing. Parameters studied were quantum - QBER, event inter-arrival jitter, event/count rate and classical - packet/flow statistics, inter-arrival mean/variance, latency proxy, TCP SYN/RST flags, byte-level Shannon entropy, with labels for benign vs. attack class. Findings: The standalone models achieved moderate performance on the held-out test set for Random Forest: ROC AUC = 0.93, F1 = 0.90, MCC = 0.86, Brier = 0.072; SVM (RBF): ROC AUC = 0.91, F1 = 0.88, MCC = 0.82, Brier = 0.081; Autoencoder (one-class): ROC AUC = 0.87, F1 = 0.83, MCC = 0.74, Brier = 0.094. By contrast, the stacked ensemble delivered perfect detection metrics for ROC AUC = 1.00, F1 = 1.00, MCC = 1.00, and Brier = 0.014. The study further emphasized the need for forensic systems to support explainability and continuous adaptability via Explainable AI and online learning with drift detection. Novelty: This study presents a cross-layer forensic framework for quantum–classical hybrid networks that fuses QKD telemetry with classical control-plane evidence and machine-learning analytics. Unlike prior work that treats these planes separately, our design unifies event-level QKD signals such QBER, arrival-time jitter with packet/flow features to produce timestamp-aligned, explainable alerts. In evaluation, the stacked-ensemble detector achieved perfect detection metrics for ROC AUC, F1, MCC and Brier on held-out data, which distinctly outperformed single-model baselines. The framework couples these gains with an XAI layer and an online, drift-aware learning loop, providing a scalable, auditable, and resilient foundation for forensic intelligence in the quantum era. Keywords: Quantum network forensics, QKD security, Advanced threat detection, Hybrid quantum-classical networks, Quantum-safe evidence, SimulaQron, Quantum cybersecurity

Thunderstorms are weather disturbances that can cause lightning, stormy winds, dense clouds, tornadoes, and heavy rain. Thunderstorms can cause extensive damage to people's lives, property, and economies, as well as livestock and national infrastructure. Early warning of thunderstorms can save people's lives and property. Previous thunderstorm prediction research did not develop a system for daily thunderstorm prediction with high accuracy for Bangladeshi citizens by assessing a wide range of meteorological variables. To address this issue, this work develops a daily high accuracy based localized thunderstorm event prediction system that analyzes various meteorological factors, dates, and specific location information. This dataset was analyzed using a variety of machine learning models, including traditional statistical models like ARMA, ARIMA, and SARIMA, as well as XGBoost ensemble methods and some deep learning models such as ANN, LSTM, and GRU. The results show that advanced neural network models, particularly GRU and LSTM, outperform others in terms of RMSE, R2, MAE, and MAPE. The GRU model outperformed all other schemes, with an RMSE of 0.794, R2 of 0.998, MAE of 0.476, and MAPE of 3.544%. The mobile application provides users with accurate, localized thunderstorm forecasts, allowing for better safety, event planning, and environmental preparedness. User feedback-based mobile app assessment confirms that more than 55% of users are highly satisfied with the thunderstorm assistance app’s features and usefulness.

Prognostic stratification of Hodgkin lymphoma (HL) patients in ICU remains challenging, with conventional scoring systems often overlooking pathophysiological biomarkers. This retrospective cohort study analyzed 1,908 HL patients from the MIMIC-IV database. Multivariate logistic regression and machine learning (ML, gradient-boosting (GBM) was optimized with LASSO regularization) were employed to identify 30-day mortality predictors, validated through SHAP interpretability, calibration curves, and decision curve analysis. Multi-organ dysfunction (AST, BUN, T-Bil), systemic inflammation (NLR, WBC) and APTT emerged as critical mortality determinants, and selected for model construction. GBM achieved superior discrimination (training AUC = 0.89; test AUC = 0.75), SHAP analysis, calibration curve and decision curve analysis (DCA) confirmed clinical utility, outperforming empirical intervention strategies. This study establishes a biomarker-driven ML framework for HL prognosis, integrating renal, hepatic, and inflammatory markers into actionable risk stratification. thereby providing a scientific basis for comprehensive HL management.

Recurrent spontaneous bacterial peritonitis (SBP) is a major concern for cirrhotic patients with ascites. This study seeks to identify predictors of recurrent SBP using clinical factors, inflammatory markers, and machine learning models. The study involved 347 patients with cirrhotic ascites and SBP. Receiver Operating Characteristic (ROC) curve analysis assessed the predictive ability of biomarkers. A composite score was created to evaluate the risk stratification model. Different machine learning models were compared for predictive accuracy. Eighty-three patients (23.9%) experienced recurrent SBP. Independent predictors of recurrence in multivariable analysis included acute kidney injury (AKI), elevated C-reactive protein (CRP) levels, higher serum bilirubin levels, a higher model for end-stage liver disease (MELD) score, proton-pump inhibitor (PPI) use, and lack of β-blocker use. A composite 10-point score (including AKI, CRP > 50mg/L, low albumin levels < 2.5g/dL, ascitic protein < 1.0g/dL, albumin/ascitic ratio < 2.5 [2 points], MELD ≥ 15, diabetes, multidrug-resistant organism [MDRO] infection, and non-use of β-blockers) stratified the risk of recurrence into low (0-3: 15%), moderate (4-6: 45%), and high (7-10: 80%) categories. Machine learning models outperformed supervised machine logistic regression in predicting recurrence. Logistic regression achieved 70% accuracy, 65% sensitivity, and 68% specificity. The decision tree model improved accuracy to 75%, sensitivity to 72%, and specificity to 71%. The random forest model showed the best performance with 78% accuracy, 77% sensitivity, and 76% specificity. A composite score, combined with machine-learning models like random forest, enhances risk assessment for SBP recurrence. Clinical predictors such as AKI, CRP, bilirubin, MELD, PPI use, and β-blockers non-use help in targeted prevention.

Latent tuberculosis infection (LTBI) is a significant reservoir for active tuberculosis development. Identifying key risk factors is crucial for prevention strategies. Machine learning techniques can uncover complex relationships between risk factors and disease outcomes. Data were collected from China's Tuberculosis Management Information System. LTBI was defined by positive tuberculin skin tests. A case-control design comparing LTBI (n = 669) with active tuberculosis (ATB, n = 669) patients was employed. Propensity score matching (1:1) was performed using age, gender, and education level. Four machine learning models (random forest, XGBoost, support vector machine, and neural network) were developed for feature importance analysis. Least Absolute Shrinkage and Selection Operator (LASSO) regression and logistic regression identified key risk factors. Bootstrap resampling (n = 1,000 iterations) assessed model stability with 95% confidence intervals. Shapley Additive Explanations (SHAP) analysis provided feature importance interpretation. A risk nomogram was constructed and evaluated using receiver operating characteristic curves, calibration plots, and decision curve analysis. Among 1,338 matched participants, XGBoost demonstrated superior performance (AUC = 0.898, accuracy = 85.7%, sensitivity = 84.2%, specificity = 86.9%). SHAP analysis revealed age group (mean |SHAP value|=0.818) as the most influential predictor, followed by medical insurance type (0.599), income group (0.523), and education level (0.439). Logistic regression identified 11 significant risk factors: age (OR = 2.35, 95%CI: 1.86-2.96), BMI (OR = 0.81, 95%CI: 0.71-0.93), smoking status, occupational dust exposure, diabetes, medical insurance type, immunosuppressant use, education level, silicosis, anemia, and TB contact history. The nomogram showed good discrimination (AUC = 0.839) and clinical utility, identifying 64.44% of subjects as high-risk with 53.62% confirmed as true positives at 20% risk threshold. This study successfully identified key LTBI risk factors using machine learning approaches. The developed nomogram provides a practical tool for targeted screening in resource-limited settings. Interventions targeting modifiable factors such as smoking cessation and occupational dust control may reduce LTBI and active TB burden.

With unmanned aerial vehicles (UAVs), agricultural monitoring has developed into a new phase of innovation providing remedies to precision farming. The common traditional agricultural methods are based on manual inspection and few observations on the ground using sensors that may be inaccurate and time-consuming. New technologies such as drones and AI provide us with an opening of large scale, early detection, but most systems currently only seek pests or diseases and are usually specific to a single type of crop in controlled laboratory conditions. Drone-operated AI system, which combines RGB and, where feasible, multispectral cameras and a YOLOv8 pipeline to detect pests and crop diseases simultaneously across a variety of crops. We are developing it to be used in the real world: we load in data fields, laboratories, and the internet, perform preprocessing, transfer learning, and make the inference to be lightweight enough to execute on edge computers. The introduction of agricultural monitoring systems based on the use of UAVs builds on the peculiarities of quadcopters and fixed-wing UAVs. Quadcopters are used when conducting detailed field surveys or spot checks, allowing high-resolution imaging to be used in order to complete precise inspections, whereas fixed-wing UAVs are used when it comes to covering extensive areas and long-range capabilities. These UAVs can gather extensive data and conduct biological and chemical analyses due to sophisticated IoT devices and sensors, such as multispectral and hyperspectral cameras, GPS modules, and real-time communication tools. Our hybrid machine learning model (HMLM) has more accuracy and predictive capabilities, with an amazing score of 98.74 and hence, our machine learning model is doing the right job of 98.74 accurate classification and thereby yielding high accurate yields by predicting crop management. This research will contribute to the sustainability of agricultural practices as well as yield protection by providing timely, precise and scalable detection. The model proposed can potentially enable farmers with action-oriented insights, losses can be alleviated, and food security objectives can be achieved in areas where there are high susceptibility rates to pests and diseases.

This study evaluates the mechanical performance and predictive modeling of fine-grained soils stabilized with crushed aggregate residue (CAR) or crushed limestone waste (CLW) and Portland cement by integrating the porosity–binder index (η/Civ) and Machine Learning (ML) techniques. Laboratory testing included unconfined compressive strength (qu) and small-strain shear modulus (Go) measurements on mixtures containing 15% and 30% CAR and 3% and 6% cement, compacted at dry unit weights between 1.69 and 1.81 g·cm−3 and cured for 7 and 28 days. Results revealed that strength and stiffness increased significantly with both cement and CAR contents. The mixture with 30% CAR and 6% cement exhibited the highest mechanical performance at 28 days (qu = 1550 kPa and Go = 6790 MPa). When mixtures are compared within the same curing period, the role of CAR and cement becomes evident. At 28 days, increasing CAR from 15% to 30% led to a moderate rise in qu (from 1390 to 1550 kPa) and Go (from 6220 to 6790 MPa). Likewise, at 7 days, increasing cement from 3% to 6% at 15% CAR produced significant gains in qu (207 to 693 kPa) and Go (2090 to 4120 MPa). The porosity–binder index showed strong correlations with qu (R2 = 0.94) and Go (R2 = 0.92). The ML models further improved accuracy, achieving R2 values of 0.99 for qu and 0.97 for Go. Although the index already performed well, the additional gain provided by ML is meaningful because it reduces prediction errors and better captures nonlinear interactions among mixture variables. This results in more reliable estimates for mix design, confirming that the combined use of η/Civ and ML offers a robust framework for predicting the behavior of soil–cement–CAR mixtures.

Alzheimer's disease (AD) remains one of the most challenging neurodegenerative disorders, with limited therapeutic options and high failure rates in clinical trials. This work developed a drug repurposing pipeline powered by a machine learning (ML) model to find possible glycogen synthase kinase-3 beta (GSK-3β) inhibitors, a crucial target in AD pathogenesis. We selected, pre-processed, and optimized a dataset of 4,087 experimentally verified GSK-3β inhibitors using dimensionality reduction and descriptor creation. The most excellent prediction performance was obtained by Random Forest (100 descriptors) out of six supervised ML algorithms that were studied (R2 = 0.8178, RMSE = 0.8118, MAE = 0.6084). Following the virtual screening of 1,616 Food and Drug Administration (FDA)-approved drugs using this refined model, many compounds with projected IC₅₀ < 500 nM were found. Docking experiments showed insightful interactions and high binding affinities with the active-site residues of GSK-3β. With the best docking score (-9.3kcal/mol), stable molecular dynamics (Average RMSD values (1000 ns): protein, 2.23 ± 0.93 Å; protein-ligand complex, 1.40 ± 0.43 Å) and long-lasting contacts with crucial residues, dolutegravir stood out among the top choices. ADMET profiling validated good pharmacokinetics and safety characteristics; however, possible hepatotoxicity needs more research. A HOMO-LUMO gap of 3.07eV was found by density functional theory (DFT) analysis, indicating robust electron transport characteristics and balanced reactivity that are favorable for protein-ligand interaction. Together, these findings show that dolutegravir is a potential repurposable option against AD and how integrative ML, docking, MD, ADMET, and quantum chemistry techniques may speed up the identification of new drugs.

We developed an OpenFOAM® application for generation of tri-periodic assemblies of fixed non-overlapping particles, intended for direct numerical simulations with body-fitted unstructured meshes. The particles can be spherical, cylindrical or spherocylindrical, with random or pre-assigned positions and orientations, and mono- or polydisperse. The assemblies are optimized for meshing with snappyHexMesh: various meshing errors are minimized by using automatically generated edge meshes, as well as by controlling the interparticle distance and tangentiality to the boundaries. Further, we provide a new pressure boundary condition which improves the accuracy of the resulting hydrodynamic forces. The available post-processing function objects are extended to also calculate stresslets (i.e., resistance to the straining motion), relevant for rheology of suspensions. The workflow is validated against available analytical and numerical data, showing excellent agreement. With our present contribution, an OpenFOAM® user is able to significantly reduce the pre-processing efforts: typically, packings of solid fractions up to 0.3 are generated in the range of a few seconds to around a minute. This allows for efficient gathering of data needed for formulation of closure laws or for developing machine learning models, relevant for industrial applications such as pneumatic conveying and fluidized beds.

Machine learning is a powerful tool for extracting meaningful patterns from large datasets and performing predictive modeling. In recent years, machine learning methods have been increasingly applied in family sciences, mental health, and educational research. This systematic review aims to evaluate how machine learning methods are used to understand the impact of family dynamics on individuals’ mental health, educational attainment, and behavioral outcomes. A comprehensive literature search was conducted in the Web of Science, PubMed, Scopus, Science Direct, Ulakbim, and TRDizin databases, and 11 studies meeting the PICOS criteria were analyzed. The reviewed studies indicate that machine learning algorithms provide strong predictions in areas such as domestic violence, depression, academic achievement, and children’s psychosocial development. In particular, Random Forest (RF), Support Vector Machines (SVM), deep learning, and natural language processing (NLP) methods have demonstrated high accuracy in predictive tasks. However, challenges related to model transparency, ethical concerns, and applicability within the family context remain among the limitations of machine learning models. Therefore, future research should focus on enhancing the interpretability of machine learning approaches, integrating them with theoretical models, and supporting their application in family sciences with more empirical studies. By doing so, machine learning techniques can be used more effectively to understand family dynamics and support individuals' mental health.

Coastal areas are dynamic, shaped by natural processes and human activities, making accurate prediction of shoreline and land use changes crucial for sustainable management. This study presents a hybrid modeling framework that combines CA-Markov and machine learning to enhance land use/land cover (LULC) and shoreline change projections in Oman's vulnerable coastal regions. Coastlines were delineated using multi-temporal Landsat images (1997-2006-2015-2024) and the Normalized Difference Water Index, while erosion and accretion rates were quantified using End Point Rate and Linear Regression Rate analyses. Results from 1997 to 2024 show substantial spatial variability, with urban localities such as Rakhyut experiencing significant erosion (-1.81m/year) and areas like Bawshar showing accretion (1.41m/year). Coastal LULC changes reveal rapid urban expansion, as seen in Muscat's built-up area, which increased from 10.31km² in 1997 to 116.41km² in 2015. Four models-CA-Markov, CA-Markov + XGBoost, CA-Markov + CART, and CA-Markov + RF-were evaluated for future LULC prediction. The hybrid CA-Markov + RF model achieved the highest predictive performance, increasing overall accuracy from 0.905 (CA-Markov) to 0.935 (CA-Markov + RF) on the test dataset, highlighting the capability of machine learning models. Projections for 2033 indicate continued urban growth, particularly in Salalah and Sohar, alongside reductions in vegetation in arid regions.

Parkinson's disease (PD) considerably impacts health-related quality of life (HRQoL) through motor and non-motor symptoms. The Parkinson's Disease Questionnaire-39 (PDQ-39) is the most widely used tool to assess HRQoL, encompassing eight dimensions and a Summary Index providing an overall score. Despite advances in machine learning (ML) for predicting disease symptoms and progression, its application to predict HRQoL across these dimensions remains underexplored. This study uses complete-case data for 478 of 861 patients from PRISM, a cross-sectional observational survey conducted in six European countries in 2018-2019. Participants were adults with PD recruited through advocacy groups and clinical centers who completed online assessments, providing data on demographics, medication, comorbidities, and disease characteristics (Tolosa et al., 2021). ML models were trained to predict PDQ-39 dimensions and Summary Index scores (0-100; higher = worse HRQoL). Features were preselected using the Boruta algorithm on the training data. Model selection was based on the lowest mean RMSE from 100 bootstrap resamples on the training set. Selected models were then retrained using 1000 bootstrap resamples for robust performance estimation. Final performance was evaluated on a held-out 20% validation set using R², MAE, and RMSE. Feature importance was assessed using permutation importance with MAE loss (100 permutations) on the held-out validation set. Factor Analysis of Mixed Data (FAMD) was used to explore patterns between non-motor symptoms and PDQ-39. Selected models: xgbTree (Summary Index; Activities of Daily Living) and gaussprPoly (all other PDQ-39 dimensions). On the validation set, Summary Index/ Cognitions showed the strongest performance with R² = 0.56/0.53, MAE = 9.60/12.39, RMSE = 12.66/16.20. Permutation feature importance ranked the Non-Motor Symptoms Questionnaire score (sum of 30 non-motor symptoms, range 0-30) as the most important predictor across all models. FAMD showed clustering of Social Support, Bodily Discomfort, and Stigma dimensions with Anxiety. Our findings demonstrate the critical role of non-motor symptoms in predicting HRQoL in patients with PD. While ML models effectively predict overall HRQoL and cognitive aspects, achieving comparable performance on other dimensions may require additional variables to reduce error. These insights emphasize comprehensive treatment strategies addressing both motor and non-motor symptoms.