Traditional Linear Regression Research Articles

A pioneering approach to deterministic rainfall forecasting for wet period in the Northern Territory of Australia using machine learning

A pioneering approach to deterministic rainfall forecasting for wet period in the Northern Territory of Australia using machine learning

Do machine learning methods solve the main pitfall of linear regression in dental age estimation?

Age estimation is crucial in forensic and anthropological fields. Teeth, are valued for their resilience to environmental factors and their preservation over time, making them essential for age estimation when other skeletal remains deteriorate. Recently, Machine Learning algorithms have been used in age estimation, demonstrating high levels of accuracy. However, their precision with respect to the trend of age estimation error, typical in some traditional methods like linear regression, has not been thoroughly investigated. To evaluate and compare the performance of frequently used Machine Learning-assisted methods against two traditional age estimation methods, linear regression and the Segmented Normal Bayesian Calibration model. Overall, 1.949 orthopantomographs from black and white South African children aged 5-14 years, with 49 % males, were evaluated. The performance of Random Forest, Support Vector Regression, K-Nearest Neighbors and the Gradient Boosting Method were compared against traditional linear regression and the Segmented Normal Bayesian Calibration model. The comparison was based on accuracy measures, including Mean Absolute Error and Root Mean Squared Error, and precision measures, including the Inter-Quartile Range of the error distribution and the slope of the estimated age error relative to chronological age. The Machine Learning methods outperformed linear regression and the Segmented Normal Bayesian Calibration models in terms of accuracy, although the differences were small. Gradient Boosting Method and Support Vector Regression achieved the highest levels of accuracy (Mean Absolute Error: 0.69 years, Root Mean Squared Error: 0.85 years). All Machine Learning methods and linear regression exhibited significant bias in residuals, whereas the Segmented Normal Bayesian Calibration model showed no significant bias. Gender-stratified analyses revealed similar results in terms of the accuracy and precision of all considered models. Although Machine Learning methods demonstrate high levels of accuracy, they may be prone to trends in error distribution when estimating dental age. Evaluating this error is crucial and should be an integral part of model performance evaluation. Future research should aim to improve accuracy while rigorously addressing systematic biases.

P-1812. Artificial Intelligence Modeling and Implementation in Antimicrobial Stewardship Program at a Tertiary Medical Center in Taiwan

Abstract Background Antimicrobial stewardship program (ASP) aim to promote appropriate antibiotic use and preventing antimicrobial resistance. Personalized tools using artificial intelligence (AI) to provide tailored antibiotic recommendations to healthcare providers may transform the way we approach antibiotic stewardship.Figure 1Rate of Multidrug-Resistant Organisms by Department Methods The study included all patients (age > 12) admitted to Chang Gung Memorial Hospital from January 1, 2005, through January 31, 2022, who had been prescribed antibiotics after admission and had microbiological evidence within the hospitalization. We conducted a linear regression equation to explore the relationship between antibiotic heterogeneity index (AHI) and multidrug-resistant organisms (MDROs) ratio in different units of hospital. We used a variety of machine learning algorithms to evaluate the incremental value of various variable groups of patients, and developed an AI model for predicting MDRO infections and antimicrobial de-escalation during hospitalization.Figure 2Value of Antibiotic Heterogeneity Index by DepartmentAHI: antimicrobial heterogenecity index Results There were 789,259 individuals fulfilled the including criteria. Figure 1 shows the incidence rates of MDROs across 19 departments, ranging from 3.2% in oncology to 10.7% in thoracic medicine. AHI for each department was ranging from 0.42 in neurosurgery to 0.74 in thoracic medicine (figure 2). Linear regression for the relationship between AHIs and MDRO incidence rates, represented a regression coefficient β of 0.179 (p-value: 0.02), indicating a statistically significant difference, with an R-squared of 0.47. Variables considered in modeling analysis of MDRO infection and antimicrobial de-escalation were shown in table 3. Random Forests with model 3 variables had the best (97%) positive predictive value (PPV) among five machine learning model in predicting MDRO infection, and the performance is much better than traditional linear regression (Figure 4). However, the best predicting model for antimicrobial de-escalation had only 43% PPV in Random Forests with model 5 variables.Table 3Variables Considered in Modeling Analyses of Multidrug-Resistant Organisms and Antibiotic De-escalation Tiered on Feasibility of Measurement Conclusion High AHI had positive correlation with high MDRO incidence. AI modeling had a great performance in predicting MDRO infection. Early predicted high risk group can be an important special population for ASP implementation.Figure 4Model Performance of Classification for Multidrug-Resistant Organisms by Input Variable Feasibility Tier Model 1 included tier 1 factors as predictor variables to predict interested outcome; model 2 included both tier1 and tier2 as predictor variables; model 3 use tier1 to tier3 as predictor variables; model 4 included tier1 to tier3 variables and plus tier 4 factors which contains continuous laboratory variables as predictor variables and model 5 included tier1 to tier3 variables and plus tier 5 factors which contains categorical laboratory variables as predictor variables. Disclosures All Authors: No reported disclosures

Estimation of Fire Counts and Fire Radiative Power Using Satellite Optical and Microwave Vegetation Indices With Random Forest Method

AbstractThe satellite microwave emissivity difference vegetation index (EDVI) has been used in previous studies to estimate FCs and FRP using traditional multivariate linear regression models. However, the nonlinear effects and contributions of numerous factors that affect forest fires cannot be disentangled by this model. Using the random forest (RF) model, this study utilized multiple EDVIs and the optical normalized difference vegetation index (NDVI) as key fuel properties to resolve the physical driving mechanisms of forest fires and to estimate the daily FCs and FRP over East Asia. The results showed that the estimated FCs and FRP were in good agreement with satellite observations, with a spatial R of 0.59 for FCs and 0.63 for FRP and a temporal R of 0.80 for FCs and 0.81 for FRP. The integration of EDVIs and NDVI into the RF model was found to improve model performance and generate overall lower systematic errors than the model without vegetation variables. Model performance was better than that in previous studies using multivariate linear regression models. In addition, EDVIs showed greater importance than NDVI. This was largely due to their daily temporal resolution that allowed EDVIs to capture forest fire dynamics in time. The combination of the RF model with satellite microwave and optical observations shows good performance and has great potential for FC and FRP estimations in global fire danger assessment.

Deep mining of megawatt large wind turbine actual operating data: exploration of Bayesian parameter fusion modeling

Abstract Real-time operation data analysis and condition monitoring of large wind turbines are crucial for efficient and safe operation of wind farms. In response to this, an accurate prediction model architecture based on the multivariate linear regression algorithm is proposed for a deeper understanding of the actual operation of large wind turbines. Comparing different prediction variables, we have confirmed that the average wind direction and average wind speed play a core role in predicting active power. The display has significantly better approximation than the traditional multiple linear regression (MLR), especially in the peak region of period-like variation. The quasi-periodic peak variation characteristics of the active power near the observed series 2000, 2500 and 3000 are realized, and the peak prediction results of B-MLR are significantly better than those of traditional MLR. Meanwhile, for RMSE and R-squared, B-MLR is significantly better than the traditional MLR algorithm in the active power prediction of wind turbines. However, the data complexity in actual applications poses a challenge to the effectiveness of the Bayesian fusion algorithm. Further optimization of the algorithm is needed to cope with the complex and variable real data environment. This study provides scientific evidence for the efficient operation, precise maintenance, and environmentally friendly design of wind turbines, promoting the continuous progress and development of wind power generation technology.

Prediction Interval Transfer Learning for Linear Regression Using an Empirical Bayes Approach

ABSTRACTCurrent literature on transfer learning has been focused on improving the predictive performance corresponding to a small dataset by transferring information to it from a larger but possibly biassed dataset. However, the transfer learning methods currently available do not allow the computation of prediction intervals, and hence, one has to rely on using either the small dataset alone or combining it with the possibly biassed dataset to obtain prediction intervals using traditional linear regression methods. In this article, we propose an Empirical Bayes approach for Prediction Interval Transfer Learning (EB‐PITL), to compute prediction intervals for transfer learning in linear regression tasks. We have proved that the Gibbs sampler associated with EB‐PITL is geometrically ergodic, so EB‐PITL can also quantify the Monte Carlo uncertainty associated with its predicted value. The efficiency of EB‐PITL against currently available methods is demonstrated using simulation studies and by analysing the Tetouan City power consumption dataset.

Multiple Environmental Variables as Covariates to Improve the Accuracy of Spatial Prediction Models for SOM on Karst Aera

ABSTRACTAims accurately predicting the spatial distribution of soil organic matter (SOM) is essential for environmental management and carbon storage estimation. However, the diversity of sources of variables poses a challenge in studying the spatial distribution of SOM. Methods in order to address this issue, we propose leveraging multiple environmental variables and employing machine learning models, specifically Lightweight gradient boosting machine learning (LightGBM) and random forest (RF), for predicting SOM spatial distribution. 128 soil samples were collected from the Caohai National Nature Reserve, and their SOM content was measured. Results the study found that the average SOM content was 36.75 g/kg. Compared to traditional linear regression models such as ordinary kriging (OK), ordinary least squares (OLS), and geographically weighted regression (GWR), the machine learning models based on nonlinear regression, LightGBM and RF, demonstrated higher cross‐validated coefficients of determination (R2) of 0.62 and 0.60, respectively, outperforming the other models. Additionally, RF exhibited lower mean absolute error (MAE) and root mean square error (RMSE), indicating higher stability and generalization capability. The spatial distribution of SOM among the models showed consistency, with higher SOM content observed in southern and near‐Caohai Lake regions and lower SOM content in northern and farther regions from Caohai Lake. Results from the Shapley additive explanations (SHAP) model highlighted agricultural land (AL), pH, and Elevation (ELV) as primary covariates influencing SOM spatial distribution. Conclusions this study provides valuable insights and support for environmental management and carbon storage estimation in the karst plateau region.

Modeling soil respiration in summer maize cropland based on hyperspectral imagery and machine learning

IntroductionSoil respiration (SR), the release of carbon dioxide (CO2) from soil due to the decomposition of organic matter and root respiration, is an important indicator for understanding agricultural carbon cycling and assessing anthropogenic impacts on the environment. Hyperspectral remote sensing offers a potential rapid, non-destructive approach for monitoring in agriculture. However, it remains uncertain whether hyperspectral remote sensing can provide an accurate and efficient method for estimating SR rate in croplands, particularly across different maize growth stages of under varying drought conditions.MethodsIn the study, we investigated the potential of combining hyperspectral remote sensing data with machine learning model (ML) to quantify SR rate in croplands. A drought field experiment was conducted, and SR and hyperspectral imagery were collected during four maize growth stages: Jointing Stage (JS), Tasseling Stage (TS), Flowering Stage (FS), and Grain Filling Stage (GFS). We compared the performance of traditional multiple linear regression (MLR) with that of an ML model (extreme gradient boosting, XGBoost), in simulating SR rate across these four growth stages.ResultsOur findings demonstrated that the simulation of the XGBoost model, utilizing soil temperature (Ts) and hyperspectral data, outperformed the MLR model. Across different growth stages, the SR simulated by the XGBoost model (R2 = 0.8103) was more reliable than that of the MLR model (R2 = 0.7451). The XGBoost model can also effectively capture the impact of drought treatments on SR.DiscussionThe XGBoost model’s tree-based structure allows it to effectively capture complex interactions and nonlinear patterns within variables, while its high sensitivity to changes in SR rates under drought conditions makes it more reliable for modeling SR across different growth stages compared to the linear-based MLR model. This study highlights the great promise of ML combined with hyperspectral imaging in predicting SR rate in croplands, which will help guide future agricultural management and environmental informatics.

Dynamic emotion intensity estimation from physiological signals facilitating interpretation via appraisal theory.

Appraisal models, such as the Scherer's Component Process Model (CPM), represent an elegant framework for the interpretation of emotion processes, advocating for computational models that capture emotion dynamics. Today's emotion recognition research, however, typically classifies discrete qualities or categorised dimensions, neglecting the dynamic nature of emotional processes and thus limiting interpretability based on appraisal theory. In our research, we estimate emotion intensity from multiple physiological features associated to the CPM's neurophysiological component using dynamical models with the aim of bringing insights into the relationship between physiological dynamics and perceived emotion intensity. To this end, we employ nonlinear autoregressive exogeneous (NARX) models, as their parameters can be interpreted within the CPM. In our experiment, emotions of varying intensities are induced for three distinct qualities while physiological signals are measured, and participants assess their subjective feeling in real time. Using data-extracted physiological features, we train intrasubject and intersubject intensity models using a genetic algorithm, which outperform traditional sliding-window linear regression, providing a robust basis for interpretation. The NARX model parameters obtained, interpreted by appraisal theory, indicate consistent heart rate parameters in the intersubject models, suggesting a large temporal contribution that aligns with the CPM-predicted changes.

Postoperative Apnea-Hypopnea Index Prediction of Velopharyngeal Surgery Based on Machine Learning.

To investigate machine learning-based regression models to predict the postoperative apnea-hypopnea index (AHI) for evaluating the outcome of velopharyngeal surgery in adult obstructive sleep apnea (OSA) subjects. A single-center, retrospective, cohort study. Sleep medical center. All subjects with OSA who underwent velopharyngeal surgery followed for 3 to 6 months were enrolled in this study. Demographic, polysomnographic, and anatomical variables were analyzed. Compared with traditional stepwise linear regression (LR) algorithm, machine learning algorithms including artificial neural network (ANN), support vector regression, K-nearest neighbor, random forest, and extreme gradient boosting were utilized to establish the regression model. Surgical success was defined as a ≥50% reduction in AHI to a final AHI of <20 events/h. A total of 152 OSA adult patients (median [interquartile range] age = 40 [35, 48] years, male/female = 136/16) were included in this study. The ANN model achieved the highest performance with a coefficient of determination (R 2) of 0.23 ± 0.05, a root mean square error of AHI of 10.71 ± 1.01 events/h, an accuracy for outcomes classification of 81.3% ± 1.2% and an area under the receiver operating characteristic of 74.6% ± 1.9%, whereas for LR model, they were 0.094 ± 0.06, 11.61 ± 0.76 events/h, 71.7% ± 1.5% and 68.8% ± 2.9%, respectively. The machine learning-based model exhibited excellent performance for predicting postoperative AHI, which is helpful in guiding patient selections and improving surgery outcomes.

Groundwater Quality Assessment and Modeling Using Water Quality Index and Mathematical Models: Case Study of Bam Province, Burkina Faso

In light of groundwater's fundamental role in providing drinking water to populations in the Sahel, its management and monitoring are vital for predicting and mitigating future crises. The goal of this study is to assess the groundwater quality in Bam province using the Water Quality Index (WQI), and to predict these indices through the application of Artificial Neural Networks (ANN) and traditional Multiple Linear Regression (MLR) techniques. In this context, a variety of physicochemical parameters such as Total Hardness (TH), pH, Electrical Conductivity (EC), calcium, magnesium, sodium, potassium, ammonium, bicarbonate, chloride, sulphate, nitrite, nitrate, phosphorus, and fluoride were collected from 154 boreholes, analysed, and used to calculate the WQIs. These parameters were employed as inputs, while the WQIs served as the output target for the models. The data were arranged in ascending order based on the indices, with 70% of the data reserved for training the models and 30% for testing. The groundwater quality in the study area, characterized by its geological heterogeneity and discontinuous fractures affecting groundwater flow, is predominantly excellent, with 95.45% of the samples having WQIs below 50. The remaining 4.55% is split between good (3.90%) and poor (0.65%) quality, with WQIs ranging from 50–100 and 100–200, respectively. In terms of predictive modeling, the ANN method provided the most accurate results, with R² (coefficient of correlation) = 0.99, RMSE (Root Mean Scare Error) = 0.0037, and MAE (Mean Absolute Error) = 0.0032 for the training set, and R² = 0.96, RMSE = 4.46, and MAE = 3.27 for the testing set. By comparison, the traditional method showed lower accuracy with R² = 0.61, RMSE = 2.71, MAE = 2.02 for the training set, and R² = 0.93, RMSE = 7.97, MAE = 6.32 for the testing set. The slight decrease in model accuracy during the testing phase is attributed to the challenge of modeling strong indices with weaker ones, as well as the geological heterogeneity and discontinuities that complicate groundwater quality prediction. However, this does not affect the model’s ability to predict extreme situations, such as water pollution events.

PREDICTION OF PERMEABILITY OF OIL AND GAS LAYERS USING ARTIFICIAL NEURAL NETWORKS

This article focuses on predicting the permeability of oil and gas reservoirs using artificial neural networks (ANN). By utilizing data sets from oil and gas wells, comprehensive preprocessing was conducted, including feature selection, scaling, and normalization to ensure the robustness of the models. The effectiveness of ANN in predicting the permeability of underground formations was evaluated using petrophysical data from wells in the Bukhara-Khiva oil and gas region. A precise permeability prediction model was created using key petrophysical parameters such as gamma rays (GR), resistivity (RT), sonic (DT), density (RHOB), and neutron porosity (NPHI). To enhance model performance, the dataset underwent complete preprocessing, including normalization and feature selection. The model's performance was assessed through MSE, R², and MAE metrics, demonstrating higher accuracy compared to traditional linear regression models. The results indicate that the ANN model provides highly accurate permeability predictions. The findings offer valuable insights for optimizing exploration and production strategies in the oil and gas industry, highlighting the superiority of machine learning and neural network models over traditional methods in subsurface resource evaluation.

Research on the Prediction of Yield and Portfolio Planning of Actively Managed Funds Based on Machine Learning

This paper takes 23 features of China's actively managed funds from 2015 to 2024 as samples, and predicts the future returns of the funds based on traditional linear regression models, Elastic Net models, Decision Tree models, Random Forest models, Ridge Regression models and Lasso Regression models, and constructs investment portfolios in groups according to the predicted return. The effectiveness of the investment portfolio construction is examined by analyzing the Sharpe ratio, information ratio, volatility and maximum drawdown rate of real data, and the predictive ability and stability of the features are examined by analyzing the value of Rank IC and ICIR of the features. The study found that the features used in this paper can provide rich information for the model, and most of the features have strong predictive ability and predictive stability; the performance of each investment portfolio is consistent with the real data; compared with the traditional linear model, the machine learning method has higher prediction accuracy, more flexibility and stability.

The impact of social features on the financial sustainability of pension systems in Central and Eastern Europe and Baltic States

This paper analyses the financial sustainability of pension systems from the perspectives of labour supply, population aging, and demographic changes. The time period chosen was 1995 to 2022 and it takes into account nine countries from Central and Eastern Europe and the Baltic States. The econometric modelling techniques used are traditional linear regression, ridge and lasso regression, cross-validated glmnet models, and xboost flexible model used for the whole nine countries and also for country groups. The empirical results show that government expenditures and revenues in terms of social security funds are influenced by the demographic changes of the population that we face nowadays, are bound to the population aging phenomenon, and are dependent upon the elderly labour supply movements. The conclusions of this article reveal the practical need to reshape the financial sustainability of pension systems in all nine countries by developing a sustainable pension scheme that needs to be adjusted to the new social, demographic, and behavioural patterns of labour supply.

Identifying Effect Modification of Latent Population Characteristics on Risk Factors with a Sparse Varying Coefficient Regression.

Leveraging observational data to understand the associations between risk factors and disease outcomes and conduct disease risk prediction is a common task in epidemiology. While traditional linear regression and other machine learning models have been extensively implemented for this task, the associations between risk factors and disease outcomes are typically deemed fixed. In many cases, however, such associations may vary by some underlying features of the individuals, which may involve certain subpopulation characteristics and environmental factors. While data for these latent features may not be available, the observed data on risk factors may have captured some proportion of the variation in these features. Thus extracting latent factors from risk factors and incorporating this effect modification into the model may better capture the underlying data structure and improve inference. We develop a novel regression model with some coefficients varying as functions of latent features extracted from the risk factors. We have demonstrated the superiority of our approach in various data settings via simulation studies. An application on a dataset for lung cancer patients from The Cancer Genome Atlas (TCGA) Program showed that our approach led to a 6% - 118% increase in (AUC-0.5) for distinguishing between different lung cancer stages compared to the classic lasso and elastic net regressions and identified interesting latent effect modifications associated with certain gene pathways.

The evaluation of building materials moisture using the Analysis of Covariance

Abstract The paper presents the evaluation of some building materials moisture using the Analysis of Covariance (ANCOVA). The readouts present the dependence between the moisture of material, evaluated gravimetrically and permittivity values determined by TDR technique. Thanks to the ANCOVA method, it is possible to determine the moisture content in the particular materials regardless of the sensor construction and type of material. Additionally, in order to compare, the data obtained in the research also were analyzed using the traditional linear regression models.

PM2.5 IoT Sensor Calibration and Implementation Issues Including Machine Learning

Affordable IoT PM2.5 sensors, enabled by the Internet of Things, offer new ways to monitor air quality. However, concerns exist about their data accuracy. This study aimed (1) to investigate the low-cost PM sensor's performance under various outdoor ambient circumstances and (2) to evaluate seven calibration methods, which include decision trees, gradient-boosted trees, linear regression, nearest neighbors, neural networks, random forests, and the Gaussian Process. The Davis AirLink was used as a reference to compare the Plantower PMS3003 sensor's performance. The data from the Plantower PMS3003 sensor were then compared to the Davis AirLink values using calibration curves created by machine learning algorithms. Calibration curves were generated using machine learning algorithms trained on sensor measurements collected in two Thai cities (Nakhon Si Thammarat and Phuket). Our results show that all machine learning methods outperformed traditional linear regression, with decision trees and neural networks demonstrating the most significant improvement. This research highlights the need for sensor calibration and the limitations of current calibration methods and paves the way for advancements in cloud-based calibration and machine learning for improved data accuracy in IoT PM2.5sensor technology. Doi: 10.28991/ESJ-2024-08-06-08 Full Text: PDF

Application of fiber-optic strain sensing technology in high-precision load prediction of aircraft landing gear

In this paper, a compact and lightweight high-precision airborne fiber-optic strain sensor is designed, and a strain-load prediction model based on convolutional neural network and long short-term memory network (ConvLSTM) is proposed to validate the collected landing gear strain data. Firstly, based on the fiber Bragg grating (FBG) sensing principle and simulation validation, small and lightweight strain and temperature sensors were fabricated, and key performance parameters such as sensitivity and linearity were comprehensively investigated, and they were mounted on the left landing gear of the aircraft to conduct strain monitoring by loading a three-way load, and the ConvLSTM model was used to train and test the strain-load mapping with the maximum relative error, average relative error and variance as indicators to evaluate its prediction accuracy and stability. The experimental results show that the strain measurement accuracy is maintained within 2.5 % and the strain sensitivity is as high as 1.3 pm/με within the ± 5000 με measurement range of the strain sensor; the maximum relative errors of the X, Y, and Z load predictions are 6.03 %, 3.75 %, and 4.12 %, respectively, and the overall average relative errors are 2.38 %, 0.27 %, and 0.76 %, with variances of 0.23 N, 0.61 N, and 0.61 N, respectively. 0.23 N, 0.61 N and 0.12 N, indicating that the model predictions are stable and highly accurate, demonstrating higher prediction accuracy when compared with traditional multiple linear regression methods. The results of this research have important application value in the field of aircraft structural health monitoring.

Surface temperature change amplifies social inequality in heat exposure? Evidence from Shanghai, China

Heat-related environmental justice has gained increasing attention, with many studies focusing on social inequalities in exposure to extreme surface temperatures. However, changes in land surface temperature (LST) also contribute to residents’ thermal discomfort, underscoring the need to explore inequalities in exposure to these changes. This study addresses this gap by analyzing spatial and social disparities in exposure to both LST changes and summer LST in Shanghai. We employed a boosted regression tree model to quantify LST changes, offering better predictive performance than traditional linear regression, and used geographically weighted regression with data from China's sixth population census to assess the city-wide and local scale inequality differences in exposure to LST changes and summer LST. Our findings reveal that 57.83 % of Shanghai's subdistricts experienced above-average LST changes and summer LST. Specific groups—males (22.17 %), children (22.61 %), the elderly (13.04 %), and residents in older housing (60.87 %)—were disproportionately exposed to local-scale LST changes. In contrast, those employed in agriculture, forestry, and fisheries faced consistently lower exposure. Furthermore, exposure to LST changes was 3.47 % higher among the elderly and 28.26 % higher for residents in older housing compared to exposure based on summer LST alone. To mitigate these inequalities, we recommend increasing water coverage and green roofs, enhancing green infrastructure in aging neighborhoods, and providing financial subsidies for outdoor workers. These findings emphasize the importance of addressing social inequalities in exposure to temperature changes to enhance urban resilience and promote sustainable urban planning in the face of climate change.

Prediction of National Innovation Using Data Scientific Approach

Countries have traditionally relied on various innovation measures to plan national-wise strategies and analysis, which hinders their ability to take proactive actions beforehand. This study introduces an advanced model for predicting national innovativeness using machine learning techniques. Utilizing a comprehensive dataset that includes 1,410 observation points and 20 variables from 141 countries spanning from 2011 to 2020, the data were sourced from the World Bank Open Data, Global Entrepreneurship Monitor (GEM) and Global Innovation Index (GII). The proposed model employs machine learning algorithms such as regression tree, random forest, support vector machine and extreme gradient boosting (XGBoost), and benchmarks their performance against the traditional linear regression analysis method. The findings reveal that all five machine learning models significantly outperform the traditional linear regression model in terms of correlation, root mean square error (RMSE) and mean absolute error (MAE), with XGBoost demonstrating superior performance. The XGBoost analysis highlights university and industry research collaboration, knowledge workers and logistics performance as the most critical variables influencing national innovativeness. This research not only presents a robust predictive model leveraging machine learning but also contributes to theoretical advancements by uncovering previously overlooked variables, offering new insights and practical implications for enhancing national innovation strategies.