Safeguarding autonomy: A focus on machine learning decision systems

Forest fires harm forests and endanger lives. Susceptibility assessment mapping using remote sensing and geographic information systems can help prevent forest fires and allocate fire-fighting resources rationally. This study combines machine learning (ML) and multi-criteria decision analysis to propose LOGEAHP – an integrated method based on logistic regression and the analytic hierarchy process (AHP). This method is used to calculate the relative weights of forest fire influencing factors and generate a forest fire susceptibility map for the Greater Khingan Range. The results show that nearly one-third of the Greater Khingan Range faced severe high forest fire risks between 2013–2024, with higher fire risks in the northern and southern regions than those in the central region. “Distance from Road” is the most critical factor affecting forest fire risks in the Greater Khingan Range, with a feature value as high as 1.1287. The LOGEAHP exhibits excellent fire assessment performance and stability, with an average receiver operating characteristic (ROC) of 0.9311 and 95% confidence interval for ROC ranging from 0.91 to 0.95. Bootstrap validation indicates that LOGEAHP achieves a statistically significant improvement over ML and AHP, with p-value significantly less than 0.01. This approach provides a robust basis for local fire prevention strategies.

This paper describes a machine learning (ML) decision support system to provide a list of chemotherapeutics that individual multiple myeloma (MM) patients are sensitive/resistant to, based on their proteomic profile. The methodology used in this study involved understanding the parameter space and selecting the dominant features (proteomics data), identifying patterns of proteomic profiles and their association to the recommended treatments, and defining the decision support system of personalized treatment as a classification problem. During the data analysis, we compared several ML algorithms, such as linear regression, Random Forest, and support vector machines, to classify patients as sensitive/resistant to therapeutics. A further analysis examined data-balancing techniques that emerged due to the small cohort size. The results suggest that utilizing proteomics data is a promising approach for identifying effective treatment options for patients with MM (reaching on average an accuracy of 81%). Although this pilot study was limited by the small patient cohort (39 patients), which restricted the training and validation of the explored ML solutions to identify complex associations between proteins, it holds great promise for developing personalized anti-MM treatments using ML approaches.

Awkward kneeling in sloped shingle installation operations exposes roofers to knee musculoskeletal disorder (MSD) risks. To address the varying levels of risk associated with different phases of shingle installation, this research investigated utilizing machine learning to automatically classify seven distinct phases in a typical shingle installation task. The classification process relied on analyzing knee kinematics data and roof slope information. Nine participants were recruited and performed simulated shingle installation tasks while kneeling on a sloped wooden platform. The knee kinematics data were collected using an optical motion capture system. Three supervised machine learning classification methods (i.e., k-nearest neighbors (KNNs), decision tree (DT), and random forest (RF)) were selected for evaluation. The KNN classifier provided the best performance for overall accuracy. The results substantiated the feasibility of applying machine learning in classifying shingle installation phases from workers’ knee joint rotation and roof slope angles, which may help facilitate method and tool development for automated knee MSD risk surveillance and assessment among roofers.

The health care process involves a set of massive data that can be of different types and placed in different places. The digitization of information, if properly analysed, leads to an increase in the quality of processes and services in the Healthcare sector. In this context, Machine Learning (ML) is the key enabling technology to reduce the costs of health care extracting knowledge from data. The goal of ML is to improve a system without any continuous human intervention. Medical Decision Support Systems (MDSS) are an added value on the one hand by speeding up diagnosis and care processes, on the other hand reducing the time of medical staff or specialized learning. The paper proposes a new method to enable medical decision making improving collaboration between machine learning experts and clinicians in order to model individual treatments. The idea is to adapt machine learning models and decision support systems to address the challenging task of the decision of individual treatments in order to make more efficient decision-making and more effective therapies.

Retrospective cohort study. The objective of this study was to determine prognostic factors for the progression of osteoporotic vertebral fracture (OVF) following conservative treatment. Few studies have evaluated factors associated with progressive collapse (PC) of OVFs. Furthermore, machine learning has not been applied in this context. The study involved the PC and non-PC groups based on a compression rate of 15%. Clinical data, fracture site, OVF shape, Cobb angle, and anterior wedge angle of the fractured vertebra were evaluated. The presence of intravertebral cleft and the type of bone marrow signal change were analyzed using magnetic resonance imaging. Multivariate logistic regression analysis was performed to identify prognostic factors. In machine learning methods, decision tree and random forest models were used. There were no significant differences in clinical data between the groups. The proportion of fracture shape ( P <0.001) and bone marrow signal change ( P =0.01) were significantly different between the groups. Moderate wedge shape was frequently observed in the non-PC group (31.7%), whereas the normative shape was most common in the PC group (54.7%). The Cobb angle and anterior wedge angle at diagnosis of OVFs were higher in the non-PC group (13.2±10.9, P =0.001; 14.3±6.6, P <0.001) than in the PC group (10.3±11.8, 10.4±5.5). The bone marrow signal change at the superior aspect of the vertebra was more frequently found in the PC group (42.5%) than in the non-PC group (34.9%). Machine learning revealed that vertebral shape at initial diagnosis was a main predictor of progressive vertebral collapse. The initial shape of the vertebra and bone edema pattern on magnetic resonance imaging appear to be useful prognostic factors for progressive collapse in osteoporotic vertebral fractures.

Reproducible Machine Learning-Based Voice Pathology Detection: Introducing the Pitch Difference Feature.

Modern cloud platforms host large-scale data pipelines that power analytics, machine learning, and real-time decision systems across industries ranging from healthcare to finance. As data volumes grow exponentially and workloads become increasingly heterogeneous spanning batch analytics, streaming ingestion, and GPU-accelerated ML tasks traditional rule-based scheduling and static resource provisioning mechanisms struggle to satisfy evolving performance, cost, and reliability constraints. These legacy approaches often rely on fixed heuristics, overprovisioned clusters, or reactive scaling, leading to suboptimal utilization, latency bottlenecks, and inflated operational expenses. This article explores AI-driven approaches to optimizing cloud data pipelines by embedding machine learning into workflow scheduling, predictive autoscaling, anomaly detection, and cross-layer resource orchestration. Grounded in foundational distributed systems research such as the MapReduce paradigm introduced by Jeffrey Dean and Sanjay Ghemawat, as well as the DAG-based dataflow model of Apache Spark, and extended by modern serverless execution engines and global optimization frameworks like AGORA, we propose an architectural blueprint that leverages telemetry-driven feedback loops, predictive modeling, and reinforcement learning to enable adaptive, cost-aware decision-making. By synthesizing key research developments from 2000 to 2022, this work outlines how intelligent scheduling policies, workload forecasting, and heterogeneity-aware optimization can transform static pipelines into self-optimizing systems, while also identifying future research directions in reinforcement learning–based orchestration, multi-cloud cost modeling, and explainable AI for cloud execution environments.

Problem definition: Human donor milk provides critical nutrition for millions of infants who are born preterm each year. Donor milk is collected, processed, and distributed by milk banks. The macronutrient content of donor milk is directly linked to infant brain development and can vary substantially across donations, which is why multiple donations are typically pooled together to create a final product. Approximately half of all milk banks in North America do not have the resources to measure the macronutrient content of donor milk, which means pooling is done heuristically. For these milk banks, an approach is needed to optimize pooling decisions. Methodology/results: We propose a data-driven framework combining machine learning and optimization to predict macronutrient content of donations and then optimally combine them in pools, respectively. In collaboration with our partner milk bank, we collect a data set of milk to train our predictive models. We rigorously simulate milk bank practices to fine-tune our optimization models and evaluate operational scenarios such as changes in donation habits during the COVID-19 pandemic. Finally, we conduct a year-long trial implementation, where we observe the current nurse-led pooling practices followed by our intervention. Pools created by our approach meet clinical macronutrient targets approximately 31% more often than the baseline, although taking 60% less recipe creation time. Managerial implications: This is the first paper in the broader blending literature that combines machine learning and optimization. We demonstrate that such pipelines are feasible to implement in a healthcare setting and can yield significant improvements over current practices. Our insights can guide practitioners in any application area seeking to implement machine learning and optimization-based decision support. History: This paper has been accepted as part of the 2023 Manufacturing &amp; Service Operations Management Practice-Based Research Competition. Supplemental Material: The e-companion is available at https://doi.org/10.1287/msom.2022.0455 .

The paper proposes a framework for optimizing and improving steam power plant operations using artificial intelligence (AI). In particular, we will do a case study on the boiler subsystem, which constitutes a significant portion of the steam power plant. Moreover, this power plant has other important subsystems such as; a condenser, a turbine, feed water, and other minor auxiliaries. The boiler subsystem will be studied and simulated using Simulink environment in MATLAB. Our objective is to utilize this simulation model to gather data that may be used in an AI approach. By contrasting our data-set with the real power plant, we may adjust our simulation model. The simulation data set was selected because it enables simple input parameter modification without involving or affecting the real power plant. Consequently, after using the machine learning model, comparing the outputs of the data from the actual power plant with the simulated data set. Furthermore, a validation technique will be used to determine how the data-set of the regression model will optimize the real power plant. This paper discusses the proposed framework and provides an explanation of the machine learning technique with which the data-set will move to a further process of training and testing using different regression analyses. Furthermore, this paper focuses on machine learning algorithms, namely, linear regression, decision tree, and random forest. In addition using r2-score for comparing the performance of these algorithms.

Cholera remains a critical public health concern in Nigeria, causing significant mortality and straining healthcare resources. Despite efforts to control the disease, recurrent outbreaks persist, especially in areas with inadequate sanitation and limited access to clean water. Traditional models for predicting cholera outbreaks often fall short due to their inability to capture the multifactorial nature of the disease's spread. This study aims to address this gap by employing regression machine learning (ML) models that integrate various socio-economic, environmental, and demographic factors to optimize the prediction of total cholera outbreak in Nigeria. The research utilized a comprehensive dataset, including variables such as population size, health facilities, water sources, poverty rates, financial inclusion, and literacy levels, spanning across 76 local government areas (LGAs) in 21 states of Nigeria. Six machine learning models Linear Regression, Decision Trees, Random Forests, XGBoost, LightGBM, and Neural Networks were developed and compared to identify the most effective predictive approach. Data preprocessing, feature engineering, and exploratory data analysis (EDA) were conducted to ensure data quality and model readiness. The results indicate that LightGBM outperforms other models, with the lowest Mean Absolute Error (MAE) and the highest R-squared value, explaining 69% of the variance in cholera cases. This model's ability to handle complex interactions between variables suggests its potential utility in real-time public health interventions. The study concludes that integrating advanced machine learning techniques into public health surveillance can significantly enhance cholera outbreak prediction and support proactive strategies to mitigate the disease's impact. These findings offer actionable insights for policymakers, public health authorities, and researchers aiming to improve disease surveillance and response in Nigeria and other developing regions.

Health is the major worrisome point whose impalpability increases with increase in the age. Thus, taking care of elders is very important responsibility. In such scenario, technology is helping people by providing living assistance. One of the major causes of health degradation or death of elders is `fall'. In this paper, a fall detection system is proposed based on machine learning. The system detects falls by classifying different activities into fall and non-fall actions and alert the relative or care taker of the elderly person in case of emergency. The dataset SisFall with variety of activities of multiple participants is used to calculate features. Machine learning algorithms SVM and decision tree are used to detect the falls on the basis of calculated features. The system acquires accuracy up to 96% by using decision tree algorithm.

The production of waste marble is the ultimate concern of the construction industry’s recent success in satisfying the huge need for marble stones. This study predicts the compressive strength of waste marble powder concrete using machine learning methods XG Boost, AdaBoost, Cat-Boost, Gradient-Boosting, Light Gradient-Boosting, and decision tree. For this purpose, a comprehensive dataset was developed from previously published literature, incorporating input features such as cement, fine and coarse aggregates, superplasticizers, silica fume, marble dust, and water, with compressive strength (CS) as the output parameter. Based on the outcomes, XG Boost model outperforms other models to predict CS with R2 values of 0.999 and 0.915 on the train and test stage, respectively. The decision tree model also shows consistent performance with R2 > 0.85 for both train and testing phases. In addition, experimental assessment was also conducted to verify the outcomes of machine learning (ML) modeling and quantify the CS of concrete specimens having marble dust as binder replacement (5–20% by weight). In addition, feature importance analysis highlights that superplasticizers, cement, and silica fume were the most weighted factors that affected CS. According to partial dependency plot (PDP) analysis, combined data on concrete strength versus cement concentration shows that the strength of the material increases from 37.5 MPa to 57.5 MPa when the cement concentration increases from 300 kg/m3 to 500 kg/m3. Furthermore, microstructural evaluation revealed a dense and well-compacted matrix in concrete containing waste marble powder, with improved bonding between the binder and aggregates.

This paper discusses a novel hybrid approach to text classification that integrates a machine learning algorithm along with DistilBERT, a pre-trained deep learning framework for natural language processing, offers a base model fine-tuned on Indian Banking News-Events with a rule-based method that is used by filtering false positives and dealing with false negatives to enhance the results given by the previous classifier. The major benefit is that by incorporating unique rules for certain chaotic or overlapping categories that have not been effectively trained, the system can be quickly fine-tuned. This research also compares the effectiveness of state-of-art deep contextual language representation DistilBERT used in our proposed hybrid model with the most preferred context independent language representation TFIDF, on supervised learning of the classification of multiclass Banking News-Events. Both representations are fed into the machine learning classifiers, Logistic Regression, Linear SVC, Decision Tree, and Random Forest. The findings indicate that DistilBERT can better transfer generic domain knowledge to other domains as compared to the baseline TFIDF and results into higher accuracy with Random Forest. The result is further fed into the Rules based method as per our proposed model. And, our proposed hybrid model resulted into improved accuracy.

Ever since the introduction of Bitcoin, cryptocurrencies have attracted interest from many due to their potential for appreciation. They have also been stigmatized for their volatility, low correlation with traditional assets, and uncertain regulatory conditions. Nevertheless, many cryptocurrencies still attract the attention of major players in finance. As in the management of a portfolio of other assets, there is a need to regularly rebalance the portfolio to optimize returns across different time horizons. This study evaluates four machine learning models, Logistic Regression, Decision Tree, K-Nearest Neighbors, and Gradient Boosting, for identifying optimal cryptocurrency rebalancing decisions. Using a momentum-based approach with a 30-day forward window, the models are trained to classify assets as hold or rebalance based on future price movement. Our approach is based on feature engineering and hyperparameter tuning. Results show that tree-based models, such as Decision Tree and Gradient Boosting, demonstrate superior classification performance in identifying optimal rebalancing moments. However, the study also highlights the limitations of using historical data exclusively without referring to other external factors such as market sentiment and regulatory changes. Overall, the study makes a contribution to the field of cryptocurrency portfolio management by providing one of the first comparative evaluations of multiple ML architectures for cryptocurrency rebalancing decisions, demonstrating the potential of machine learning to improve portfolio management in highly volatile markets.

Chronic fatigue severely compromises the quality of life in patients with granulomatosis with polyangiitis (GPA). Traditional diagnostic methods are often time-consuming, relying on clinical expertise and detailed questionnaires. This study aimed to develop a machine learning model capable of predicting chronic fatigue in GPA patients based on clinical data, with a particular focus on improving diagnostic capacity in regions with limited access to specialists. This cross-sectional study collected data on fatigue (measured by the Modified Fatigue Impact Scale, MFIS), functional ability (Health Assessment Questionnaire, HAQ), disease activity (Birmingham Vasculitis Activity Score, BVAS), comorbidities, medication use, physical activity (International Physical Activity Questionnaire - Short Form, IPAQ-SF), and demographic characteristics. Four machine learning algorithms-logistic regression, decision tree, random forest, and extreme gradient boosting-were assessed using a 70/30 train-test split. Model performance was evaluated using area under the curve (AUC), accuracy, F1 score, recall, and precision. Statistical comparisons were performed using Welch's t-test and the Wilcoxon-Mann-Whitney U test for continuous variables, while the chi-square test or Fisher's exact test was applied to categorical variables, with significance set at P < 0.05. All analyses were conducted using R version 4.4.1 for Windows. Forty-five patients were assessed: 62.2% were female, with a median BMI of 27.72kg/m² (23.2-30.1), a median age of 55.5 years, and a median disease duration of 12.0 years (6.0-17.0). Fatigue was reported by 20 patients (MFIS score ≥ 38), and seven patients (15.5%) had active disease according to the BVAS, which was similar between the fatigued and no fatigued groups (P > 0.05). The fatigued group had more acute-phase reactants and prednisone use (P < 0.05). The tree-based models achieved an AUC of approximately 0.80, outperforming the other models. Tree-based models demonstrated superior predictive performance in identifying chronic fatigue. The Random Forest model, in particular, highlighted higher disability in activities of daily living (HAQ), older age, and longer disease duration as key predictors. Although the models performed well, additional data and incorporation of clinically relevant variables may further enhance predictive accuracy. Patients with GPA who experienced fatigue showed higher glucocorticoid use and elevated acute-phase reactants, despite similar levels of disease activity, suggesting mechanisms beyond inflammation. Machine learning shows strong potential as a clinical tool for fatigue identification, especially in settings with limited access to specialist care. Universal Trial Number (UTN): U1111-1271-6003; Brazilian Clinical Trials Registry (ReBEC): RBR-9n4z2hh. Registration date: January 18, 2022, and Plataforma Brasil (CAAE # 41762820.1.0000.0068).