Search Results (198)

Traditional machine learning models for concrete compressive strength prediction provide only single-value estimates without quantifying the probability of meeting design requirements, leaving engineers unable to make risk-informed decisions. This study addresses this critical limitation by developing a novel probabilistic prediction framework that integrates explainable machine learning with Copula-based joint distribution modeling. Using a dataset of 1030 concrete samples with curing ages ranging from 1 to 365 days, we first established an XGBoost 2.1.4 prediction model achieving R² = 0.9211 (RMSE = 4.51 MPa) on the test set. SHAP 0.49.1 (SHapley Additive exPlanations) analysis identified curing age (33.3%) and water–cement ratio (28.8%) as the dominant features, together accounting for 62.1% of predictive importance. These two controllable engineering parameters were then selected as core variables for probabilistic modeling. The key innovation lies in integrating Copula-based dependence modeling with explainable machine learning (XGBoost–SHAP) to quantify the compliance probability of concrete strength under specific mix designs and curing conditions, thereby supporting risk-informed quality control decisions. Through systematic comparison of five Copula families (Gaussian, Student t, Clayton, Gumbel, and Frank), we identified optimal dependence structures: Gaussian Copula (ρ = −0.54) for the water–cement ratio–strength relationship and Clayton Copula for the age–strength relationship, revealing asymmetric tail dependence patterns invisible to conventional correlation analysis. The three-dimensional Copula model enables engineers to estimate compliance probability—the likelihood of concrete achieving target strength under specific mix designs and curing conditions. We propose an illustrative three-tier decision rule for construction quality management based on the compliance probability P: P ≥ 0.95 (high-confidence approval), 0.80 ≤ P < 0.95 (warning zone requiring enhanced monitoring), and P < 0.80 (high risk suggesting corrective actions such as mix adjustment or extended curing), noting that these thresholds can be recalibrated to project-specific risk tolerance and local specifications. This framework supports a paradigm shift from reactive “mix-then-test” quality control to proactive “predict-then-decide” construction management, providing quantitative risk assessment tools previously unavailable in deterministic prediction approaches. Full article

The recent use of large language models (LLMs) in code generation and code summarization tasks has been widely adopted by the software engineering community. New LLMs are emerging regularly with improved functionalities, efficiency, and expanding data that allow models to learn more effectively. The lack of guidelines for selecting the right LLMs for coding tasks makes the selection a subjective choice by developers rather than a choice built on code complexity, code correctness, and linguistic similarity analysis. This research investigates the use of machine learning classification and ranking methods to select the best-suited open-source LLMs for code generation and code summarization tasks. This work conducts a comparison experiment on four open-source LLMs (Mistral, CodeLlama, Gemma 2, and Phi-3) and uses the MBPP coding question dataset to analyze code-generated outputs in terms of code complexity, maintainability, cyclomatic complexity, code structure, and LLM perplexity by collecting these as a set of features. An SVM classification problem is conducted on the highest correlated feature pairs, where the models are evaluated through performance metrics, including accuracy, area under the ROC curve (AUC), precision, recall, and F1 scores. The RankNet ranking methodology is used to evaluate code summarization model capabilities by measuring ROUGE and BERTScore accuracies between LLM code-generated summaries and the coding questions used from the dataset. The study results show a maximum accuracy of 49% for the code generation experiment, with the highest AUC score reaching 86% among the top four correlated feature pairs. The highest precision score reached is 90%, and the recall score reached up to 92%. Code summarization experiment results show Gemma 2 scored a 1.93 RankNet win probability score, and represented the highest ranking reached among other models. The phi3 model was the second-highest ranking with a 1.66 score. The research highlights the potential of machine learning to select LLMs based on coding metrics and paves the way for advancements in terms of accuracy, dataset diversity, and exploring other machine learning algorithms for other researchers. Full article

The rapid development of rural and peri-urban areas increases the demand for decentralized wastewater treatment systems. Small wastewater treatment plants (SWTPs) with a capacity below 2000 PE are becoming an important element of local water protection and circular-economy strategies, yet clear guidelines for selecting appropriate technologies are still lacking. This study analyzes the criteria used in decision-making for SWTPs from a multi-stakeholder perspective and evaluates the relative importance of technical, economic, environmental and social factors. The research was conducted in Poland and included a survey of 130 respondents representing six stakeholder groups (officials, operators, designers, contractors, scientists and residents). Respondents allocated weights to four main groups of criteria and assessed eleven detailed parameters on a 1–10 scale. The data were analyzed using descriptive statistics, the Kolmogorov–Smirnov test with the Lilliefors correction to verify distribution assumptions, and the Kruskal–Wallis test to examine differences between stakeholder groups. The results show a consistent hierarchy of criteria, with technical reliability, treatment efficiency and operating costs ranked as the most important factors. Social and environmental aspects were assessed as relevant but secondary. Only minor differences between stakeholder groups were observed. The study highlights the need for integrated, multicriteria approaches in SWTP planning, particularly in dispersed rural areas. The findings may support local authorities, designers and investors in technology selection. The research is limited by the non-probability sampling strategy, the national scope of the dataset and the cross-sectional character of the survey. Full article

Background: Early mortality risk stratification is essential in acute pulmonary embolism (PE). Integrating clinical variables with biomarkers may enhance prognostic accuracy beyond established tools. Methods: In a retrospective cohort of 322 patients with confirmed acute PE, we evaluated syncope, right-ventricular (RV) dysfunction, systolic blood pressure (SBP) and biochemical markers as candidate predictors of in-hospital mortality. A penalized logistic regression model using LASSO (least absolute shrinkage and selection operator) was developed and internally validated with five-fold cross-validation and 200 bootstrap repetitions. Discrimination, calibration and clinical utility were assessed using the area under the receiver operating characteristic curve (AUC), Brier score and decision-curve analysis (DCA). Results: In-hospital mortality was 5.6% (n = 18). LASSO retained four predictors: syncope, RV dysfunction, lower SBP and higher troponin levels. The optimism-corrected AUC was 0.70 (95% CI 0.63–0.77), with strong calibration (Brier score 0.066). DCA showed that the model provided greater net benefit than treat-all, treat-none, and sPESI strategies across threshold probabilities of approximately 7–25%, supporting its potential value for early triage. NT-proBNP, D-dimer and lactate did not add incremental predictive information after penalization. Conclusions: A simple, interpretable model integrating clinical parameters and troponin demonstrates good predictive performance and favorable clinical utility for early mortality risk estimation in acute PE. External validation is required before broader implementation. Full article

Objective: In this study, we aimed to develop and internally validate a clinically applicable nomogram for predicting live birth following in vitro fertilization (IVF) using routinely available clinical and embryological parameters. Methods: This retrospective study was conducted at a single tertiary IVF center. Women undergoing IVF/ICSI were included if their baseline demographic and clinical data were available, they had undergone at least one fresh or frozen–thawed embryo transfer, and they had a known live birth outcome. Women with cycles without embryo transfer and those missing key outcome data were excluded from the analysis. As a result, a total of 2119 IVF/ICSI treatment cycles resulting in embryo transfer were included in the analysis. To identify independent predictors of live birth, multivariable logistic regression analysis was performed. Results: Among the 2119 treatment cycles analyzed, 541 resulted in live birth (25.5%). Multivariable logistic regression with backward stepwise selection identified female age (OR: 0.959, p < 0.001), high embryo quality (OR: 2.752, p < 0.001), day of embryo transfer (day 5 vs. day 3, OR: 1.427, p = 0.001), and endometrial thickness on the day of transfer as independent predictors of live birth (OR: 1.086, p < 0.001). These variables were incorporated into a nomogram (the Zübeyde Hanim IVF Nomogram) to estimate individualized live birth probability. The model demonstrated acceptable discrimination, with a bootstrap-corrected area under the receiver operating characteristic curve (AUC) of 0.64 (95%CI: 0.61–0.66), and it showed satisfactory calibration across deciles of predicted risk. Conclusions: The Zubeyde Hanim IVF Nomogram provides an individualized and clinically practical tool for predicting live birth following IVF treatment. Based on routinely available parameters, this model may assist clinicians in patient counseling and treatment planning. Full article

In this paper, we propose FedENLC, an end-to-end noisy label correction model that performs model training and label correction simultaneously to fundamentally mitigate the label noise problem of federated learning (FL). FedENLC consists of two stages. In the first stage, the proposed model employs Symmetric Cross Entropy (SCE), a robust loss function for noisy labels, and label smoothing to prevent the model from being biased by incorrect information in noisy environments. Subsequently, a Bayesian Gaussian Mixture Model (BGMM) is utilized to detect noisy clients. BGMM mitigates extreme parameter bias through its prior distribution, enabling stable and reliable detection in FL environments where data heterogeneity and noisy labels coexist. In the second stage, only the top noisy clients with high noise ratios are selectively included in the label correction process. The selection of top noisy clients is determined dynamically by considering the number of classes, posterior probabilities, and the degree of data heterogeneity. Through this approach, the proposed model prevents performance degradation caused by incorrect detection, while improving both computational efficiency and training stability. Experimental results show that FedENLC achieves significantly improved performance over existing models on the CIFAR-10 and CIFAR-100 datasets under data heterogeneity settings along with four noise settings. Full article

This study examines the determinants of goodwill impairment recognition under IFRS 3 in the context of Moroccan listed firms. Using an unbalanced panel covering the period of 2006–2024 and comprising 862 firm-year observations, we employ a three-stage empirical strategy that integrates a Probit model to estimate the likelihood of impairment, a Tobit model to assess the magnitude of the loss, and a Heckman two-step procedure to correct for potential self-selection. The results show that goodwill impairment reflects key economic and financial fundamentals, including revenue growth, book-to-market ratios, and operating performance. However, both real and accrual-based earnings management significantly influence the probability and intensity of impairment, particularly through abnormal cash flows and income-smoothing behavior. Discretionary accruals become significant only after correcting for selection bias, indicating that they do not drive the recognition decision but contribute to determining the size of the impairment once it has been recorded. The findings are robust across multiple specifications and contribute to the broader literature on financial reporting quality under IAS/IFRS, while enriching empirical evidence on managerial discretion and earnings management in emerging-market environments. Full article

This study presents an optimization-driven design of a wireless communications network to continuously transmit environmental variables—temperature, humidity, weight, and water usage—in poultry farms. The reference site is a four-shed facility in Quito, Ecuador (each shed $120\mathrm{m}\times 12\mathrm{m}$) with a data center located $200\mathrm{m}$ from the sheds. Starting from a calibrated log-distance path-loss model, coverage is declared when the received power exceeds the receiver sensitivity of the selected technology. Gateway placement is cast as a mixed-integer optimization that minimizes deployment cost while meeting target coverage and per-gateway capacity; a capacity-aware greedy heuristic provides a robust fallback when exact solvers stall or instances become too large for interactive use. Sensing instruments are Tekon devices using the Tinymesh protocol (IEEE 802.15.4g), selected for low-power operation and suitability for elongated farm layouts. Model parameters and technology presets inform a pre-optimization sizing step—based on range and coverage probability—that seeds candidate gateway locations. The pipeline integrates MATLAB R2024b and LpSolve 5.5.2.0 for the optimization core, Radio Mobile for network-coverage simulations, and Wireshark for on-air packet analysis and verification. On the four-shed case, the algorithm identifies the number and positions of gateways that maximize coverage probability within capacity limits, reducing infrastructure while enabling continuous monitoring. The final layout derived from simulation was implemented onsite, and end-to-end tests confirmed correct operation and data delivery to the farm’s data center. By combining technology-aware modeling, optimization, and field validation, the work provides a practical blueprint to right-size wireless infrastructure for agricultural monitoring. Quantitatively, the optimization couples coverage with capacity and scales with the number of endpoints M and candidate sites N (binaries $M+N+MN$). On the four-shed case, the planner serves 72 environmental endpoints and 41 physical-variable endpoints while keeping the gateway count fixed and reducing the required link ports from 16 to 4 and from 16 to 6, respectively, corresponding to optimization gains of up to 82% and 70% versus dense baseline plans. Definitions and a measurement plan for packet delivery ratio (PDR), one-way latency, throughput, and energy per delivered sample are included; detailed long-term numerical results for these metrics are left for future work, since the present implementation was validated through short-term acceptance tests. Full article

Nanosatellites in Low Earth Orbit (LEO) are an attractive platform for commercial and scientific missions, but their downlink capacity is constrained by bandwidth and by low ground station duty cycles (often under 5%). These limitations are particularly acute in heterogeneous cooperative networks, where operators seek to maximize “good-put”: the number of unique messages successfully delivered to the ground. In this paper, we present and evaluate three complementary algorithms for scheduling nanosatellite passes to maximize good-put under realistic traffic and link variability. First, a Cooperative Reception Algorithm uses Shapley value analysis from cooperative game theory to estimate each station’s marginal contribution (considering signal quality, geography, and historical transmission patterns) and prioritize the most valuable upcoming satellite passes. Second, a pair-utility optimization algorithm refines these assignments through local, pairwise comparisons of reception probabilities between neighboring stations, correcting selection biases and adapting to changing link conditions. Third, a weighted bidding algorithm, inspired by the Helium reward model, assigns a price per message and allocates passes to maximize expected rewards in non-commercial networks such as SatNOGS and TinyGS. Simulation results show that all three approaches significantly outperform conventional scheduling strategies, with the Shapley-based method providing the largest gains in good-put. Collectively, these algorithms offer a practical toolkit to improve throughput, fairness, and resilience in next-generation nanosatellite communication systems. Full article

This paper proposes a novel multiscale random forest model for stock index trend prediction, incorporating statistical inference principles to improve classification confidence. Traditional random forest classifiers rely on majority voting, which can yield biased estimates of class probabilities, especially under small sample sizes. To address this, we introduce a multiscale bootstrap correction mechanism into the ensemble framework, enabling the estimation of third-order accurate approximately unbiased p-values. This modification replaces naive voting with statistically grounded decision thresholds, improving the robustness of the model. Additionally, stepwise regression is employed for feature selection to enhance generalization. Experimental results on CSI 300 index data demonstrate that the proposed method consistently outperforms standard classifiers, including standard random forest, support vector machine, and weighted k-nearest neighbors model, across multiple performance metrics. The contribution of this work lies in the integration of hypothesis testing techniques into ensemble learning and the pioneering application of multiscale bootstrap inference to financial time series forecasting. Full article

Climate change is increasing the drought frequency and severity, so projecting spatiotemporal drought evolution across climate zones is critical for drought mitigation. Model biases, the choice of drought index, and neglecting CO₂ effects on potential evapotranspiration (PET) add large uncertainties to future drought projections. We selected 10 global climate models (GCMs) that participated in the Coupled Model Intercomparison Project Phase 6 and downscaled model outputs using the bias correction and spatial downscaling (BCSD) method. We then developed a CO₂-aware standardized moisture anomaly index (SZI[CO₂]) and used a three-dimensional drought identification method to extract the duration, area, and severity; we then analyzed their spatiotemporal dynamics. To account for nonstationarity, Copula-based approaches were used to estimate joint drought probabilities with time-varying parameters. Projections indicate wetting in Southern Northwest China, Inner Mongolia, and the Western Tibetan Plateau (reduced drought frequency, duration, intensity), while Central and Southern China show a drying trend in the 21st century. Three-dimensional drought metrics exhibit strong nonstationarity; nonstationary log-normal and generalized extreme value distributions fit most regions best. Under equal drought characteristic values, co-occurrence probabilities are higher under SSP5-8.5 scenarios than SSP2-4.5 scenarios, with the largest scenario differences over the Tibetan Plateau and Central and Southern China. Full article

Background and Objectives: The growing academic interest within the biomedical sciences regarding the diagnostic capabilities of multimodal language models, such as ChatGPT-4o, is clear. However, their ability to interpret oral clinical images remains insufficiently explored. This exploratory pilot study aimed to provide preliminary observations about the diagnostic validity of ChatGPT-4o in identifying oral squamous cell carcinoma (OSCC), oral leukoplakia (OL), and oral lichen planus (OLP) using only clinical photographs, without the inclusion of additional clinical data. Materials and Methods: Two general dentists selected 23 images of oral lesions suspected to be OSCC, OL, or OLP. ChatGPT-4o was asked to provide a probable diagnosis for each image on 30 occasions, generating a total of 690 responses. The responses were then evaluated against the reference diagnosis set up by an expert to calculate sensitivity, specificity, predictive values, and the area under the ROC curve. Results: ChatGPT-4o demonstrated high specificity across the three conditions (97.1% for OSCC, 100% for OL, and 96.1% for OLP), correctly classifying 90% of OSCC cases (AUC = 0.81). However, this overall accuracy was largely driven by correct negative classifications, while the clinically relevant sensitivity for OSCC was only 65%. In spite of that, sensitivity was highly variable: 60% for OL and just 25% for OLP, which limits its usefulness in a clinical setting for ruling out these conditions. The model achieved positive predictive values of 86.7% for OSCC and 100% for OL. Given the small dataset, these findings should be interpreted only as preliminary evidence. Conclusions: ChatGPT-4o demonstrates potential as a complementary tool for the screening of OSCC in clinical oral images. Nevertheless, the pilot nature of this study and the reduced sample size highlight that larger, adequately powered studies (with several hundred cases per pathology) are needed to obtain robust and generalizable results. Nevertheless, its sensitivity remains insufficient, as a significant proportion of true cases were missed, underscoring that the model cannot be relied upon as a standalone diagnostic tool. Full article

We propose a sequential procedure with a closed and adaptive structure. It selects a subset of size $t(>0)$ from $k(\ge t)$ treatments in such a way that any treatment superior to the control is guaranteed to be included. All the experimental treatments and the control are assumed to produce two binary endpoints, and the procedure is based on those two binary endpoints. A treatment is considered superior if both its endpoints are larger than those of the control. While responses across treatments are assumed to be independent, dependence between endpoints within each treatment is allowed and modeled via an odds ratio. The proposed procedure comprises explicit sampling, stopping, and decision rules. We demonstrate that, for any sample size n and parameter configuration, the probability of correct selection remains unchanged when switching from the fixed-sample-size procedure to the sequential one. We use the bivariate binomial and multinomial distributions in the computation and derive design parameters under three scenarios: (i) independent endpoints, (ii) dependent endpoints with known association, and (iii) dependent endpoints with unknown association. We provide tables with the sample size savings achieved by the proposed procedure compared to its fixed-sample-size counterpart. Examples are given to illustrate the procedure. Full article

Burned area (BA) mapping and fire severity assessment are essential for understanding fire occurrence patterns, formulating post-fire restoration strategies and evaluating vegetation recovery processes. However, existing BA datasets are primarily derived from coarse-resolution satellite imagery and often lack sufficient consideration of fire severity. To address these limitations, this study utilized dense time-series Landsat imagery available on the Google Earth Engine, applying the qualityMosaic method to generate annual composites of minimum normalized burn ratio values. These composites imagery enabled the rapid identification of fire sample points, which were subsequently used to train a random forest classifier for estimating per-pixel burn probability. Pixels with a burned probability greater than 0.9 were selected as the core of the BA, and used as candidate seeds for region growing to further expand the core and extract complete BA. This two-stage extraction method effectively balances omission and commission errors. To avoid the repeated detection of unrecovered BA, this study developed distinct correction rules based on the differing post-fire recovery characteristics of forests and grasslands. The extracted BA were further categorized into four fire severity levels using the delta normalized burn ratio. In addition, we conducted a quantitative validation of the BA mapping accuracy based on Sentinel-2 data between 2015 and 2023. The results indicated that the BA mapping achieved an overall accuracy of 93.90%, with a Dice coefficient of 82.04%, and omission and commission error rates of 26.32% and 5.25%, respectively. The BA dataset generated in this study exhibited good spatiotemporal consistency with existing products, including MCD64A1, FireCCI51, and GABAM. The BA fluctuated significantly between 1985 and 2010, with the highest value recorded in 1987 (13,315 km²). The overall trend of BA showed a decline, with annual burned areas remaining below 2000 km² after 2010 and reaching a minimum of 92.8 km² in 2020. There was no significant temporal variation across different fire severity levels. The area of high-severity burns showed a positive correlation with the annual total BA. High-severity fire-prone zones were primarily concentrated in the northeastern, southeastern, and western parts of the study area, predominantly within grasslands and forest–grassland ecotone regions. Full article

Introduction: Decision-making regarding radiotherapy techniques for patients with nasopharyngeal cancer requires a comparison of photon and proton plans generated using planning software, which requires time and expertise. We developed a fully automated decision tool to select patients for proton therapy that predicts proton therapy (XT) and photon therapy (PT) dose distributions using only patient CT image data, predicts xerostomia and dysphagia probability using predicted critical organ mean doses, and makes decisions based on the Netherlands’ National Indication Protocol Proton therapy (NIPP) to select patients likely to benefit from proton therapy. Methods: This study used 48 nasopharyngeal patients treated at the Cancer Hospital of the Chinese Academy of Medical Sciences. We manually generated a photon plan and a proton plan for each patient. Based on this dose distribution, photon and proton dose prediction models were trained using deep learning (DL) models. We used the NIPP model to measure xerostomia levels 2 and 3, dysphagia levels 2 and 3, and decisions were made according to the thresholds given by this protocol. Results: The predicted doses for both photon and proton groups were comparable to those for manual plan (MP). The Mean Absolute Error (MAE) for each organ at risk in the photon and proton plans did not exceed 5% and showed a good performance of the dose prediction model. For proton, the normal tissue complication probability (NTCP) of xerostomia and dysphagia performed well, p > 0.05. There was no statistically significant difference. For photon, the NTCP of dysphagia performed well, p > 0.05. For xerostomia p < 0.05 but the absolute deviation was 0.85% and 0.75%, which would not have a great impact on the prediction result. Among the 48 patients’ decisions, 3 were wrong, and the correct rate was 93.8%. The area under curve (AUC) of operating characteristic curve (ROC) was 0.86, showing the good performance of the decision-making tool in this study. Conclusions: The decision tool based on DL and NTCP models can accurately select nasopharyngeal cancer patients who will benefit from proton therapy. The time spent generating comparison plans is reduced and the diagnostic efficiency of doctors is improved, and the tool can be shared with centers that do not have proton expertise. Trial registration: This study was a retrospective study, so it was exempt from registration. Full article