Traditional Statistical Models Research Articles

A finger on the pulse of cardiovascular health: estimating blood pressure with smartphone photoplethysmography-based pulse waveform analysis

Smartphone photoplethysmography (PPG) offers a cost-effective and accessible method for continuous blood pressure (BP) monitoring, but faces persistent challenges with accuracy and interpretability. This study addresses these limitations through a series of strategies. Data quality was enhanced to improve the performance of traditional statistical models, while SHapley Additive exPlanations (SHAP) analysis ensured transparency in machine learning models. Waveform features were analyzed to establish theoretical connections with BP measures, and feature engineering techniques were applied to enhance prediction accuracy and model interpretability. Bland–Altman analysis was conducted, and the results were compared against reference devices using multiple international standards to evaluate the method's feasibility. Data collected from 127 participants demonstrated strong correlations between smartphone-derived digital waveform features and those from reference BP devices. The mean absolute errors (MAE) for systolic BP (SBP), diastolic BP (DBP), and pulse pressure (PP) using multiple linear regression models were 7.75, 6.35, and 4.49 mmHg, respectively. Random forest models further improved these values to 7.34, 5.79, and 4.45 mmHg. Feature importance analysis identified key contributions from time-domain, frequency-domain, curvature-domain, and demographic features. However, Bland–Altman analysis revealed systematic biases, and the models barely meet established accuracy standards. These findings suggest that while smartphone PPG technology shows promise, significant advancements are required before it can replace traditional BP measurement devices.

Predictive ability of visit-to-visit glucose variability on diabetes complications

BackgroundIdentification of prognostic factors for diabetes complications are crucial. Glucose variability (GV) and its association with diabetes have been studied extensively but the inclusion of measures of glucose variability (GVs) in prognostic models is largely lacking. This study aims to assess which GVs (i.e., coefficient of variation (CV), standard deviation (SD), and time-varying) are better in predicting diabetic complications, including cardiovascular disease (CVD), diabetic retinopathy (DR), and chronic kidney disease (CKD). The model performance between traditional statistical models (adjusting for covariates) and machine learning (ML) models were compared.MethodsA retrospective cohort of type 2 diabetes (T2D) patients between 2010 and 2019 in Ramathibodi Hospital was created. Complete case analyses were used. Three GVs using HbA1c and fasting plasma glucose (FPG) were considered including CV, SD, and time-varying. Cox proportional hazard regression, ML random survival forest (RSF) and left-truncated, right-censored (LTRC) survival forest were compared in two different data formats (baseline and longitudinal datasets). Adjusted hazard ratios with 95% confidence intervals were used to report the association between three GVs and diabetes complications. Model performance was evaluated using C-statistics along with feature importance in ML models.ResultsA total of 40,662 T2D patients, mostly female (61.7%), with mean age of 57.2 years were included. After adjusting for covariates, HbA1c-CV, HbA1c-SD, FPG-CV and FPG-SD were all associated with CVD, DR and CKD, whereas time-varying HbA1c and FPG were associated with DR and CKD only. The CPH and RSF for DR (C-indices: 0.748–0.758 and 0.774–0.787) and CKD models (C-indices: 0.734–0.750 and 0.724–0.740) had modestly better performance than CVD models (C-indices: 0.703–0.730 and 0.698–0.727). Based on RSF feature importance, FPG GV measures ranked higher than HbA1c GV, and both GVs were the most important for DR prediction. Both traditional and ML models had similar performance.ConclusionsWe found that GVs based on HbA1c and FPG had comparable performance. Thus, FPG GV may be used as a potential monitoring parameter when HbA1c is unavailable or less accessible.

Optimal subsampling for double generalized linear models with heterogeneous massive data

. With the development of information technology, massive data under heterogeneous characteristics are generated in the economic, financial, and other fields. Traditional statistical models and existing statistical methods are often inadequate for handling dispersion modeling problems with heterogeneous massive data. In this article, the optimal subsampling of double generalized linear models is studied in heterogeneous massive data environments. Under certain conditions, the optimal subsampling probabilities of the double generalized linear models with heterogeneous data are derived based on the A-optimality criterion and L-optimality criterion, respectively. Furthermore, a two-step algorithm based on uniform sampling is developed, and the asymptotic properties of the subsample estimator from this algorithm are discussed. The results of numerical simulations and a real example show that the algorithm can improve estimation accuracy and decrease computational costs to some extent.

Performance Evaluation of Long Short-Term Memory and Autoregressive Integrated Moving Average Time Series Models for Stock Price Prediction

Abstract: Stock price prediction is a critical task in financial markets, often complicated by the challenges of modeling highly volatile, non-linear, and dynamic time series data. This study evaluates the performance of two prominent forecasting models: Long Short-Term Memory (LSTM) networks, known for their ability to capture long-term dependencies and non-linear patterns, and the Auto-Regressive Integrated Moving Average (ARIMA), a traditional statistical model adept at linear trend modeling. Historical stock price data from the Nigerian Exchange Limited (NGX) was utilized. Both models were implemented in a Python environment, and their predictive accuracy was assessed using performance metrics, including Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and R-squared (R²). Additionally, the Diebold-Mariano (DM) test was employed to statistically compare the models’ predictive accuracies. This study highlights the potential of LSTM models for robust stock price forecasting and provides valuable insights for selecting predictive models based on data complexity and market dynamics.

Retail Sales Prediction

Accurate retail sales forecasting is critical for optimizing inventory management, reducing operational costs, and improving strategic decision-making. Traditional statistical models often fail to capture complex consumer behaviors and market fluctuations, leading to inaccurate predictions. With advancements in machine learning (ML), predictive models now offer enhanced accuracy and adaptability in forecasting sales trends. This study evaluates the effectiveness of machine learning algorithms, specifically k-Nearest Neighbor (kNN) regression, Multinomial regression, and Decision Tree with AdaBoost regression, in predicting retail sales. The research assesses model performance using standard evaluation metrics, highlighting the advantages and limitations of each approach. The findings provide valuable insights for retailers, business strategists, and policymakers, offering a data-driven framework for improving sales forecasting and decision-making processes. Keywords: Retail Sales, Forecasting, Machine Learning, Regression Models, Predictive Analytics, Business Intelligence

Accuracy of machine learning and traditional statistical models in the prediction of postpartum haemorrhage: a systematic review

ObjectivesTo evaluate whether postpartum haemorrhage (PPH) can be predicted using both machine learning (ML) and traditional statistical models.DesignDiagnostic systematic review and meta-analysis of observational and clinical studies, prospectively registered on...

Machine Learning-Based Early Warning of Algal Blooms: A Case Study of Key Environmental Factors in the Anzhaoxin River Basin

Algal blooms are a major risk to aquatic ecosystem health and potable water safety. Traditional statistical models often fail to accurately predict algal bloom dynamics due to their complexity. Machine learning, adept at managing high-dimensional and non-linear data, provides a superior predictive approach to this challenge. In this study, we employed support vector machine (SVM), random forest (RF), and backpropagation neural network (BPNN) models to predict the severity of algal blooms in the Anzhaoxin River Basin based on an algal density-based grading standard. The SVM model demonstrated the highest accuracy with training and test set accuracies of 0.96 and 0.92, highlighting its superiority in small-sample learning. The Shapley Additive Explanations (SHAP) technique was utilized to evaluate the contribution of environmental variables in various predictive models. The results show that TP is the most significant environmental factor affecting the algal bloom outbreak in Anzhaoxin River, and the phosphorus management strategy is more suitable for the management of the artificial water body in northeast China. This study contributes to exploring the potential application of machine learning models in diagnosing and predicting riverine ecological issues, providing valuable insights and support for the protection and management of aquatic ecosystems in the Anzhaoxin River Basin.

Machine learning or traditional statistical methods for predictive modelling in perioperative medicine: A narrative review.

Machine learning or traditional statistical methods for predictive modelling in perioperative medicine: A narrative review.

Risk Scoring Models in Healthcare Data Analytics: A Practitioner’s Perspective on Machine Learning, Interpretability, and Regulatory Compliance

The development of risk scoring models has significantly contributed to the field of healthcare analytics because of its capabilities in predicting diseases, managing healthcare resources, and even affecting policy formulation. Advanced machine learning (ML) techniques have been developed over time that are outpacing the ability of traditional statistical models to achieve their goals. However, concerns regarding model transparency, real-world validation, and compliance with regulatory bodies, chiefly in relation to HIPAA and HCC in the US, persist. This paper presents consolidated comparisons of testimonies acquired during the rollout of an enterprise scale risk scoring module leveraging patient health records and claims documents focusing on effective model design, interpretability, and compliance imperatives. The objective of this paper is to provide other healthcare practitioners and researchers who wish to employ sophisticated risk scoring systems which can function seamlessly within the bounds of legal compliance and sufficient explanation afterwards with both compliance and practical aspects in mind.

Predicting Implied Volatility Via Improved Pinns–Informer to the BS Equation

This paper explores various methods for analyzing and predicting data through traditional statistical models, machine learning, and deep learning techniques. A detailed overview of SARIMA, a widely-used time series model, highlights its compo-nents-Autoregressive (AR), Moving Average (MA), Differencing, and Seasonal Differencing-that are crucial for handling time-based data dependencies. The paper also discusses machine learning approaches, such as Decision Trees and Random Forests, with an enhancement using Particle Swarm Optimization (PSO) for featureselection to improve performance. In the realm of deep learning, Convolutional Neural Networks (CNNs) are explored for their ability to handle complex data such as images, focusing on convolutional, pooling, and fully connected layers. Further, the paper presents a Physics-Informed Neural Network (PINN) model, which incorporates a dynamic loss function and attention mechanisms for improving model generalization. Finally, the hybrid model, combining deep learning with meta-learning, optimizes loss functions for faster training convergence and increased adaptability to market data. These approaches collectively aim to enhance prediction accuracy and computational efficiency across various domains.

A novel approach to the relation of multi-pollutant effect and kidney dysfunction: data analysis from the Korean National Environmental Health Survey Cycle 3 (2015-2017).

Traditional statistical models for estimating the impact of multiple environmental chemicals on kidney outcomes have limitations. This study aimed to evaluate the risk prediction of kidney disease in the general population using innovative methodologies. Serum persistent organic pollutant (POP), urinary chemical, serum creatinine, and urinary albumin levels were measured in a subpopulation of adults (n = 1,266) drawn from the Korean National Environmental Health Survey Cycle 3 (n = 3,787). Various machine learning (ML) models, including bagging, ridge, lasso, and random forest, were used to predict chronic kidney disease (CKD) risk, and their results were compared with those of conventional logistic regression methods. Furthermore, the weighted quantile sum (WQS) approach, which assigns weights to mixture components, was employed to evaluate multi-pollutant effects. Presplit was attempted to incorporate existing domain knowledge. A total of 42 variables, including baseline characteristics and laboratory findings, were analyzed during the ML modeling process. The decision tree algorithm generally outperformed logistic regression in risk prediction. Based on the decision tree models, lipid-corrected polychlorinated biphenyl 153 (PCB153) emerged as the strongest predictor of CKD. PCB153 remained a significant predictor of CKD in middle-aged adults (<50 years; p = 0.01) following age stratification. Particularly among middle-aged adults with hemoglobin levels >13.25 g/dL, CKD risk was predicted to be 71.4% in the high serum PCB153 group. Current observations showed that utilizing both WQS regression and ML-based predictions offers valuable insights. In the models, POPs, particularly PCB153, were identified as important risk factors for CKD in Korean adults.

Machine learning models for reinjury risk prediction using cardiopulmonary exercise testing (CPET) data: optimizing athlete recovery

BackgroundCardiopulmonary Exercise Testing (CPET) provides detailed insights into athletes’ cardiovascular and pulmonary function, making it a valuable tool in assessing recovery and injury risks. However, traditional statistical models often fail to leverage the full potential of CPET data in predicting reinjury. Machine learning (ML) algorithms offer promising capabilities in uncovering complex patterns within this data, allowing for more accurate injury risk assessment.ObjectiveThis study aimed to develop machine learning models to predict reinjury risk among elite soccer players using CPET data. Specifically, we sought to identify key physiological and performance variables that correlate with reinjury and to evaluate the performance of various ML algorithms in generating accurate predictions.MethodsA dataset of 256 elite soccer players from 16 national and top-tier teams in Iran was analyzed, incorporating physiological variables and categorical data. Several machine learning models, including CatBoost, SVM, Random Forest, and XGBoost, were employed to predict reinjury risk. Model performance was assessed using metrics such as accuracy, precision, recall, F1-score, AUC, and SHAP values to ensure robust evaluation and interpretability.ResultsCatBoost and SVM exhibited the best performance, with CatBoost achieving the highest accuracy (0.9138) and F1-score (0.9148), and SVM achieving the highest AUC (0.9725). A significant association was found between a history of concussion and reinjury risk (χ² = 13.0360, p = 0.0015), highlighting the importance of neurological recovery in preventing future injuries. Heart rate metrics, particularly HRmax and HR2, were also significantly lower in players who experienced reinjury, indicating reduced cardiovascular capacity in this group.ConclusionMachine learning models, particularly CatBoost and SVM, provide promising tools for predicting reinjury risk using CPET data. These models offer clinicians more precise, data-driven insights into athlete recovery and risk management. Future research should explore the integration of external factors such as training load and psychological readiness to further refine these predictions and enhance injury prevention protocols.

HGATGS: Hypergraph Attention Network for Crop Genomic Selection

Many important plants’ agronomic traits, such as crop yield, stress tolerance, and other traits, are controlled by multiple genes and exhibit complex inheritance patterns. Traditional breeding methods often encounter difficulties in dealing with these traits due to their complexity. However, genomic selection (GS), which utilizes high-density molecular markers across the entire genome to facilitate selection in breeding programs, excels in capturing the genetic variation associated with these traits. This enables more accurate and efficient selection in breeding. The traditional crop genome selection model, based on statistical methods or machine learning models, often treats samples as independent entities while neglecting the abundance latent relational information among them. Consequently, this limitation hampers their predictive performance. In this study, we proposed a novel crop genome selection model based on hypergraph attention networks for genomic prediction (HGATGS). This model incorporates dynamic hyperedges that are designed based on sample similarity to validate the efficacy of high-order relationships between samples for phenotypic prediction. By introducing an attention mechanism, it assigns weights to different hyperedges and nodes, thereby enhancing the ability to capture kinship relationships among samples. Additionally, residual connections are incorporated between hypergraph convolutional layers to further improve model stability and performance. The model was validated on datasets for multiple crops, including wheat, corn, and rice. The results showed that HGATGS significantly outperformed traditional statistical methods and machine learning models on the Wheat 599, Rice 299, and G2F 2017 datasets. On Wheat 599, HGATGS achieved a correlation coefficient of 0.54, a 14.9% improvement over methods like R-BLUP and BayesA (0.47). On Rice 299, HGATGS reached 0.45, a 66.7% increase compared to other models like R-BLUP and SVR (0.27). On G2F 2017, HGATGS attained 0.88, slightly surpassing other models like R-BLUP and BayesA (0.87). We conducted ablation experiments to compare the model’s performance across three datasets, and found that the model integrating hypergraph attention and residual connections performed optimally. Subsequent comparisons of the model’s prediction performance with dynamically selected different k values revealed optimal performance when K = (3,4). The model’s prediction performance was also compared across different single nucleotide polymorphisms (SNPs) and sample sizes in various datasets, with HGATGS consistently outperforming the comparison models. Finally, visualizations of the constructed hypergraph structures showed that certain nodes have high connection densities with hyperedges. These nodes often represent varieties or genotypes with significant impacts on traits. During feature aggregation, these high-connectivity nodes contribute significantly to the prediction results and demonstrate better prediction performance across multiple traits in multiple crops. This demonstrates that the method of constructing hypergraphs through correlation relationships for prediction is highly effective.

The Relationships Between Knowledge Hiding, Organizational Justice, Workplace Friendship and Job Interdependence

This study was designed to determine the effects of social groups and relationships, which are among the basic subjects of sociology, on management science. In this context, the relationship between knowledge hiding, organizational justice, workplace friendship and job interdependence was investigated. This research, using quantitative analysis techniques, was conducted in a public institution operating in the transportation sector. Data were collected with the help of survey forms and at this stage, a simple random sampling method was preferred without any restrictions. Data obtained from 274 employees working in TR1, TR2, TR5, TR7 and TRB regions were analyzed with traditional statistical methods and structural equation modeling (SEM). According to the findings, the tendency to hide knowledge increases in environments where organizational justice is low, the friendship bonds of employees with a high perception of organizational justice are stronger, and this reduces their knowledge hiding behaviors. The study also found that the perception of organizational justice is higher in individuals with higher work experience, organizational justice and knowledge hiding behaviors do not differ according to gender, and the tendency to hide information in blue-collar employees is much lower than in white and gray collar employees.

Discussion on “Assessing Predictability of Environmental Time Series With Statistical and Machine Learning Models” by Bonas et al.

ABSTRACTMotivated by empirical case studies and discussions of Bonas et al. (2024), this discussion paper critically examines challenges in the predictability of environmental processes, focusing on three key spheres: (a) predictability and interpretability, (b) predictability in dynamic environments, and (c) predictability into unknown spaces. These spheres highlight the responsibilities within environmetrics to ensure that predictive models, particularly advanced machine learning and deep learning methods, are applied thoughtfully. First, we discuss the trade‐off between interpretability and predictive complexity, contrasting the transparency of traditional statistical models with the “black‐box” nature of machine learning but also highlighting their enormous potential for exploiting new data sources and types. Second, we address real‐time adaptability, where models must handle concept drift and should, therefore, be continuously monitored. Finally, we consider the challenges of extrapolating predictions into unknown/nontrained areas, underscoring the risks of model overreach. This paper aims to contribute to the discussion in the field, emphasizing the critical role environmetricians play in advancing responsible, interpretable, and scientifically sound predictive practices.

From Data to Decision: Leveraging Machine Learning for Crisis Response

This study explores the integration of machine learning (ML) techniques in early warning models (EWMs) for financial crises, emphasizing decision-making in policy contexts. By com paring traditional statistical models such as logistic regression and advanced ML techniques, including boosting methods such as AdaBoost, Gradient Boosting, XGBoost, LightGBM, and CatBoost, the research evaluates predictive accuracy and decision-making efficiency. Utilizing a dataset spanning 13 countries over 49 years, this paper highlights key economic indicators such as Account-to-GDP, Inflation, and Housing Price Cycles as critical predictors. The findings underscore the superior performance of boosting models and provide actionable insights for policymakers on optimizing thresholds (τ) and balancing predictive error through relative preference (μ). Specifically, the analysis demonstrates how varying τ and μ influences model effectiveness, highlighting the trade-offs between Type I and Type II errors. This research contributes to enhancing financial stability through informed crisis anticipation and proactive policy interventions

A Comprehensive Review of Wind Power Prediction Based on Machine Learning: Models, Applications, and Challenges

Wind power prediction is essential for ensuring the stability and efficient operation of modern power systems, particularly as renewable energy integration continues to expand. This paper presents a comprehensive review of machine learning techniques applied to wind power prediction, emphasizing their advantages over traditional physical and statistical models. Machine learning methods, especially deep learning approaches such as Convolutional Neural Networks (CNNs), Long Short-Term Memory Networks (LSTMs), and ensemble learning techniques like XGBoost, excel in addressing the nonlinearity and complexity of wind power data. The review also explores critical aspects such as data preprocessing, feature selection strategies, and model optimization techniques, which significantly enhance prediction accuracy and robustness. Challenges such as data acquisition difficulties, complex terrain influences, and sensor quality issues are examined in depth, with proposed solutions discussed. Additionally, the paper highlights future research directions, including the potential of multi-model fusion, emerging deep learning technologies like Transformers, and the integration of smart sensors and IoT technologies to develop intelligent, automated, and reliable prediction systems. By addressing existing challenges and leveraging advanced machine learning techniques, this work provides valuable insights into the current state of wind power prediction research and offers strategic guidance for enhancing the applicability and reliability of prediction models in practical scenarios.

Application of Convolutional Neural Networks and Recurrent Neural Networks in Food Safety.

This review explores the application of convolutional neural networks (CNNs) and recurrent neural networks (RNNs) in food safety detection and risk prediction. This paper highlights the advantages of CNNs in image processing and feature recognition, as well as the powerful capabilities of RNNs (especially their variant LSTM) in time series data modeling. This paper also makes a comparative analysis in many aspects: Firstly, the advantages and disadvantages of traditional food safety detection and risk prediction methods are compared with deep learning technologies such as CNNs and RNNs. Secondly, the similarities and differences between CNNs and fully connected neural networks in processing image data are analyzed. Furthermore, the advantages and disadvantages of RNNs and traditional statistical modeling methods in processing time series data are discussed. Finally, the application directions of CNNs in food safety detection and RNNs in food safety risk prediction are compared. This paper also discusses combining these deep learning models with technologies such as the Internet of Things (IoT), blockchain, and federated learning to improve the accuracy and efficiency of food safety detection and risk warning. Finally, this paper mentions the limitations of RNNs and CNNs in the field of food safety, as well as the challenges in the interpretability of the model, and suggests the use of interpretable artificial intelligence (XAI) technology to improve the transparency of the model.

Explainable Self-Supervised Dynamic Neuroimaging Using Time Reversal.

Functional magnetic resonance imaging data pose significant challenges due to their inherently noisy and complex nature, making traditional statistical models less effective in capturing predictive features. While deep learning models offer superior performance through their non-linear capabilities, they often lack transparency, reducing trust in their predictions. This study introduces the Time Reversal (TR) pretraining method to address these challenges. TR aims to learn temporal dependencies in data, leveraging large datasets for pretraining and applying this knowledge to improve schizophrenia classification on smaller datasets. We pretrained an LSTM-based model with attention using the TR approach, focusing on learning the direction of time in fMRI data, achieving over 98 % accuracy on HCP and UK Biobank datasets. For downstream schizophrenia classification, TR-pretrained weights were transferred to models evaluated on FBIRN, COBRE, and B-SNIP datasets. Saliency maps were generated using Integrated Gradients (IG) to provide post hoc explanations for pretraining, while Earth Mover's Distance (EMD) quantified the temporal dynamics of salient features in the downstream tasks. TR pretraining significantly improved schizophrenia classification performance across all datasets: median AUC scores increased from 0.7958 to 0.8359 (FBIRN), 0.6825 to 0.7778 (COBRE), and 0.6341 to 0.7224 (B-SNIP). The saliency maps revealed more concentrated and biologically meaningful salient features along the time axis, aligning with the episodic nature of schizophrenia. TR consistently outperformed baseline pretraining methods, including OCP and PCL, in terms of AUC, balanced accuracy, and robustness. This study demonstrates the dual benefits of the TR method: enhanced predictive performance and improved interpretability. By aligning model predictions with meaningful temporal patterns in brain activity, TR bridges the gap between deep learning and clinical relevance. These findings emphasize the potential of explainable AI tools for aiding clinicians in diagnostics and treatment planning, especially in conditions characterized by disrupted temporal dynamics.

Analysis of Stock Prediction M+odel Based on LSTM

As the complexity and volatility of the stock market continue to increase, investors are placing greater emphasis on the need for accurate stock price predictions. Traditional statistical models often face limitations in time series forecasting, particularly when it comes to capturing the intricate and dynamic changes in stock prices. This paper explores the application of the LSTM model for stock price prediction by developing a forecasting model based on LSTM. First, relevant stock data is collected through a web crawler, and then an LSTM model is trained using this data to predict the price of a specific stock. To assess the performance of the model, evaluation metrics such as mean squared error (MSE), mean absolute error (MAE), and the coefficient of determination (R) are employed. The results demonstrate that the LSTM neural network model effectively predicts nonlinear stock trends.