Enhancing random forest algorithm performance using non-concave penalization techniques

Predicting Post-Heart Transplant Composite Renal Outcome Risk in Adults: A Machine Learning Decision Tool

Random forest estimation of genomic breeding values for disease susceptibility over different disease incidences and genomic architectures in simulated cow calibration groups

Stunting remains a major public health challenge in Indonesia, with national prevalence among children under five reaching 21.6% in 2022, according to the Ministry of Health. This condition, defined by the World Health Organization as a height-for-age less than -2 SD, is associated with long-term consequences including impaired cognitive development, reduced educational attainment, and diminished economic productivity. Addressing stunting is therefore critical to achieving Sustainable Development Goals (SDGs) related to hunger, health, and education. Despite multiple national initiatives, early identification of stunting risk is still limited by reliance on conventional, reactive surveillance methods. Recent advances in machine learning (ML) provide promising alternatives for proactive stunting prediction, with several studies reporting high predictive accuracy using ensemble methods, hybrid frameworks, and geographically weighted models. Building upon this evidence, the present study develops and evaluates ML models for stunting risk prediction using a large dataset of 10,000 records from North Sumatra, Indonesia. The dataset included three predictor variables—age, height, and weight—and a target variable, nutritional status (Normal, Stunted, Severely Stunted, Tall). Four algorithms were compared: K-Nearest Neighbors (KNN), Naïve Bayes, Decision Tree, and Random Forest. Performance was assessed using accuracy, precision, recall, F1-score, and ROC area, with 10-fold cross-validation ensuring robust estimation. Results demonstrated that Decision Tree (88.6% accuracy) and Random Forest (88.3% accuracy) outperformed KNN (84.7%) and Naïve Bayes (72%). ROC areas further confirmed the superiority of ensemble-based approaches, particularly Random Forest (0.979). Statistical significance was tested using McNemar’s test, revealing that Decision Tree and Random Forest achieved comparable performance (p = 0.651), both significantly outperforming KNN and Naïve Bayes (p < 0.05). This study contributes a context-specific evaluation of ML methods for stunting prediction in North Sumatra, emphasizing not only predictive accuracy but also interpretability to support health policy and program implementation. By bridging data-driven insights with actionable decision support, the proposed framework advances progress toward SDG-aligned strategies and provides a foundation for more targeted and preventive interventions in child nutrition and growth monitoring.

As a renewable and bio‐based material, polylactic acid (PLA) can provide long‐term sustainable environmental benefits. This study focused on analyzing the economic performance of PLA production from corn stover at a large scale in the U.S. A modeling framework is established that integrates techno‐economic analysis with the process model developed in Aspen Plus and Monte Carlo simulation that analyzes the corresponding uncertainty and variability across the system. Monte Carlo simulation is enabled by artificial intelligence (AI) models, namely Artificial Neural Network (ANN) and Random Forest. This established framework aims at streamlining the process that requires high computational time and developing an easy‐access analysis tool. Scenario analysis investigated different energy options and biorefinery lifespans. The results indicate that the mean minimum selling price of 1 t PLA is $2523 (5th percentile–95th percentile (P5–P95) $2 107–$2 940) for a 30‐year lifespan using natural gas as the energy source, $2,731 (P5–P95 $2 400–$3 076) for a 10‐year lifespan using natural gas for extra energy, and $2 637 (P5–P95 $2 370–$2 903) for a 30‐year lifespan using wood pellets for additional energy. Both AI models demonstrate the ability to predict economic performance by showing relatively high prediction accuracy, but the ANN model exhibits better performance.

The mechanism of dam safety monitoring model is analyzed; for the dam system comprehensive affected by multi-factor, the mapping relationship between the influence factors and the dam behavior effects domain is usually nonlinear. Synthesizing each kind of factor, 27 parameters are chosen as the main factors which affect the accuracy of the monitoring model. Taking the actual monitoring data as the evaluation factor, the dam safety monitoring model based on the random forest (RF) intelligent algorithm was built with the actual monitoring data to predict uplift pressure. At the same time, test the significance of each variable based on the RF monitoring model and calculate the importance degree of each variable for the model through the importance function. It is indicated that RF model can be relatively fast and accurately predict the uplift pressure of the dam according to the influence factors. The average prediction accuracy is more than 95%. As compared with other intelligent algorithms such as support vector machine, RF has better robustness, higher prediction accuracy, and faster convergence speed. Because of the uniformity of the calculation procedure and the universality of the prediction method, the RF model also has reasonable extrapolation for other dam safety monitoring models (such as crack opening and seepage discharge). Significance test results obtained by the two methods have shown that the impact of reservoir water level and daily rainfall on the uplift pressure is significant, and other factors’ impact on dam deformation is unstable and changes with the external environmental influence.

Prediction analysis of customer purchase behavior is an interesting and challenging task in modern-day life. Our objective is to introduce the concept of machine learning using a random forest algorithm in depth. In this paper, a model has been proposed for predicting which cloud services have been purchased on a number of factors. A random forest model is built using different parameters such as advertisement click sequence, previously purchased cloud services, etc. and training our model. For the execution of this proposed model, an advertisement log dataset has been taken and the necessary modifications have been done on it. As an outcome, for a customer, this proposed model is providing high accuracy in prediction. Five factors have been considered which affect purchasing decision-making of customers in cloud services, such as previous purchasing habits, a sequence of advertisements viewed, customer location, etc..

Introduction. The machine learning algorithms are highly promising for predicting the load-bearing capacity of the building structures. The paper aims at building the predictive models for calculating the strength of the concrete-filled steel tubular (CFST) columns to enable a highly accurate prediction of the ultimate loads for the entire possible range of parameters affecting the load-bearing capacity of the eccentrically compressed columns.Materials and Methods. The article studies the eccentrically compressed short concrete-filled steel tubular (CFST) columns of circular cross-section. Model input parameters: column outer diameter, pipe wall thickness, yield strength of steel, compressive strength of concrete, relative eccentricity. Output parameters: the ultimate loads without taking into account and taking into account the random eccentricities. The models were trained on synthetic data generated based on the theoretical principles of the limit equilibrium method. Two machine learning models were built. When training the first model, the ultimate loads were determined at a given eccentricity of the longitudinal force without taking into account the additional random eccentricity. When training the second model, the additional random eccentricity was taken into account. The effect of the features on the model predictions was assessed using the Feature Importance function. The Optuna method was used to select the hyperparameters. The machine learning models were implemented in the Jupiter Notebook environment using the Gradient Boosting learning method. The total volume of the training sample was 179 025 samples.Results. The importance of the features most affecting the predictive values of the model have been determined. For both models, the outer diameter of the column and the relative eccentricity have proved to be the most important features, which is consistent with the existing experience of designing and calculating such structures. Optimisation of the hyperparameters using the Grid Search method enabled getting the improved results. The high accuracy of prediction has been ascertained by the low values of the regression metrics: MSE = 9.024; MAE = 9.250; MAPE = 0.004 — for the model built without taking into account the additional random eccentricity; MSE = 8.673; MAE = 8.673; MAPE = 0.004 — for the model built taking into account the additional random eccentricity.Discussion and Conclusion. The developed Gradient Boosting models for predicting the ultimate loads of the eccentrically compressed short concrete-filled steel tubular (CFST) columns of circular cross-section, both without taking into account and taking into account the additional random eccentricities, have demonstrated high accuracy and stability of prediction, they can be applied for assessing the strength of the columns during design and construction, which will reduce the time and resources involved in physical testing. In the future, it is planned to expand the data range by including other materials, different cross-section geometries of the columns and a slenderness parameter, which may improve the generalization ability of the model.

The evaluation of soil samples using spectral techniques provides a sustainable approach to soil health assessment, reducing reliance on traditional, waste-producing analytical methods. Over the past few decades, spectroscopic and spectrometric techniques have gained prominence in soil analysis due to their non-destructive nature. X-ray fluorescence (XRF) and mid-infrared spectroscopy (MIRS) are particularly valuable, as they provide complementary information on the elemental and molecular composition of soils, respectively. Both have been successfully combined with machine learning algorithms to model and predict soil fertility parameters as alternatives to conventional wet chemistry. This study explores the potential of data fusion between XRF and MIRS measurements to enhance soil fertility prediction accuracy. A total of 160 soil samples were analyzed using a Panalytical Epsilon 5 EDXRF spectrometer, employing four different secondary targets, and a Bruker Alpha II Fourier-transform mid-infrared spectrometer. Three machine learning models were trained on individual and fused datasets: Partial Least Squares Regression (PLSR), Support Vector Machine (SVM), and Random Forest Regression (RF). The regression models built upon the fused data yielded increased performance for organic Carbon (OC), exchangeable Calcium (exCa), and exchangeable Potassium (exK). For OC, the RF model yielded the best performance, with the fused approach achieving a 5% reduction in RMSE and a 7% increase in RPD relative to standalone XRF. For exCa, RF was again the top-performing algorithm under fusion, providing a 25% reduction in RMSE and a 51% increase in RPD. For exK, the best results were obtained with PLS, which delivered a 16% reduction in RMSE and a 27% increase in RPD. These results demonstrate that integrating complementary spectral information from XRF and MIRS can enhance the prediction of key soil fertility attributes, offering a reliable and sustainable alternative to conventional chemical analyses. Beyond improving model accuracy, the proposed fusion framework highlights the potential of combining multi-sensor data to expand the applicability of spectral techniques for large-scale, rapid soil fertility assessment.

Random forest solar power forecast based on classification optimization

In the last two decades, machine learning (ML) methods have been widely used in digital soil mapping (DSM), but the regression kriging (RK) model which combines the advantages of the ML and kriging methods has rarely been used in DSM. In addition, due to the limitation of a single-model structure, many ML methods have poor prediction accuracy in undulating terrain areas. In this study, we collected the SOC content of 115 soil samples in a hilly farming area with continuous undulating terrain. According to the theory of soil-forming factors in pedogenesis, we selected 10 topographic indices, 7 vegetation indices, and 2 soil indices as environmental covariates, and according to the law of geographical similarity, we used ML and RK methods to mine the relationship between SOC and environmental covariates to predict the SOC content. Four ensemble models-random forest (RF), Cubist, stochastic gradient boosting (SGB), and Bayesian regularized neural networks (BRNNs)-were used to fit the trend of SOC content, and the simple kriging (SK) method was used to interpolate the residuals of the ensemble models, and then the SOC and residual were superimposed to obtain the RK prediction result. Moreover, the 115 samples were divided into calibration and validation sets at a ratio of 80%, and the tenfold cross-validation method was used to fit the optimal parameters of the model. From the results of four ensemble models: RF performed best in the calibration set (R2c = 0.834) but poorly in the validation set (R2v = 0.362); Cubist had good accuracy and stability in both the calibration and validation sets (R2c = 0.693 and R2v = 0.445); SGB performed poorly (R2c = 0.430 and R2v = 0.336); and BRNN had the lowest accuracy (R2c = 0.323 and R2v = 0.282). The results showed that the R2 of the four RK models in the validation set were 0.718, 0.674, 0.724, and 0.625, respectively. Compared with the ensemble models without superimposed residuals, the prediction accuracy was improved by 0.356, 0.229, 0.388, and 0.343, respectively. In conclusion, Cubist has high prediction accuracy and generalization ability in areas with complex topography, and the RK model can make full use of trends and spatial structural factors that are not easy to mine by ML models, which can effectively improve the prediction accuracy. This provides a reference for soil survey and digital mapping in complex terrain areas.

By predicting ERα bioactivity and mining the potential relationship between Absorption, Distribution, Metabolism, Excretion, Toxicity (ADMET) attributes in drug research and development, the development efficiency of specific drugs for breast cancer will be effectively improved and the misjudgment rate of R&D personnel will be reduced. The quantitative prediction model of ERα bioactivity and classification prediction model of Absorption, Distribution, Metabolism, Excretion, Toxicity properties were constructed. The prediction results of ERα bioactivity were compared by XGBoot, Light GBM, Random Forest and MLP neural network. Two models with high prediction accuracy were selected and fused to obtain ERα bioactivity prediction model from Mean absolute error (MAE), mean squared error (MSE) and R2. The data were further subjected to model-based feature selection and FDR/FPR-based feature selection, respectively, and the results were placed in a voting machine to obtain Absorption, Distribution, Metabolism, Excretion, Toxicity classification prediction model. In this study, 430 molecular descriptors were removed, and finally 20 molecular descriptors with the most significant effect on biological activity obtained by the dual feature screening combined optimization method were used to establish a compound molecular descriptor prediction model for ERα biological activity, and further classification and prediction of the Absorption, Distribution, Metabolism, Excretion, Toxicity properties of the drugs were made. Eighty variables were selected by the model ExtraTreesClassifier Classifie, and 40 variables were selected by the model GradientBoostingClassifier to complete the model-based feature selection. At the same time, the feature selection method based on FDR/FPR is also selected, and the three classification models obtained by the two methods are placed into the voting machine to obtain the final model. The experimental results showed that the model‘s evaluation indexes and roc diagram were excellent and could accurately predict ERα bioactivity and Absorption, Distribution, Metabolism, Excretion, Toxicity properties. The model constructed in this study has high accuracy, fast convergence and robustness, has a very high accuracy for Absorption, Distribution, Metabolism, Excretion, Toxicity and ERα classification prediction, has bright prospects in the biopharmaceutical field, and is an important method for energy conservation and yield increase in the future.

Aiming at improving the prediction accuracy of offshore short-term wind speed, a model based on random cuckoo search algorithm (RCSA) and artificial neural network (ANN) was proposed. Firstly, RCSA was obtained by introducing a random factor to improve CSA, and then a RCSA-ANN model for predicting offshore short-term wind speed was established. Secondly, a wind tower was built in Luchao Port, Shanghai, offshore meteorological data were measured, and the training of the model was carried out. Finally, the precision of RCSA-ANN model was verified by comparison and analysis with BP-ANN and CSA-ANN models. The results show that the improved CSA method is simple, reliable and effective, which solves the problem that the algorithm is easy to fall into local optimum. The average error of RCSA-ANN model is not only lower than that of BP-ANN model, but also much lower than that of CSA-ANN model, and the prediction accuracy of these three models decreases in turn. RCSA-ANN model has high prediction accuracy and can precisely predict fluctuating wind speed sequences, and it also has good application potential.

In today's rapidly evolving industrial landscape, efficient predictive maintenance solutions are essential for minimizing downtime and enhancing productivity. This research introduces an adaptive cloud-based model pipeline for predicting the Remaining Useful Life (RUL) of machine components, specifically ball screws. The pipeline integrates local preprocessing, edge computing, and cloud-based adaptive model training, ensuring data privacy and reducing data transmission volumes. The system classifies wear states using various machine learning mod-els and predicts RUL through regression analysis, incorporating uncertainty quantification for robust maintenance scheduling. The experimental setup includes accelerated degradation of ball screws, with data collected via a three-dimensional accelerometer. Feature extraction and data augmentation techniques are employed to enhance prediction accuracy. Random Forest and Gradient Boosting models demonstrate superior performance, with Random Forest selected for its robustness and uncertainty quantification capabilities. Empirical results indicate high prediction accuracy, with Random Forest achieving up to 91% accuracy in Phase 2. This cloud-ready predictive maintenance framework leverages scalable cloud infrastructure for efficient data processing and real-time updates, offering a practical solution for industrial applications. The proposed approach significantly advances the adoption of digital business models within the manufacturing industry, providing a reliable and efficient tool for predictive maintenance.

In the era of smart cities and personalized services, recommendation systems are an integral part of handling massive user-generated content and providing personalized recommendations. Traditional systems often fail to capture the nuances of embedded emotions and sentiment in user reviews, crucial to enhancing user engagement and satisfaction. Thereby, we propose the Emotion and Sentiment-Enriched Decision Transformer (ESE-DT), a novel framework that integrates sentiment and emotion analysis into sequential recommendation modeling. Based on the state-of-the-art sequence modeling framework of Decision Transformer (DT) originally developed for reinforcement learning, ESE-DT reimagines recommendation tasks as trajectory modeling problems. By treating user interactions as sequences of states, actions, and rewards, DT allows ESE-DT to learn temporal dynamics, enabling alignment between recommendations and long-term user engagement goals. This is further enhanced by the advanced NLP for sentiment and emotion extraction to improve the contextual understanding in user and item embeddings. We experimentally validate the proposed approach on two real-world datasets, namely, Yelp and Google Local Reviews. Results clearly show that ESE-DT outperforms state-of-the-art baselines along several key metrics, achieving up to 11.76% improvement in nDCG@10, 11.58% increase in HR@10, and a 42.11% reduction in RMSE on the Yelp dataset, while attaining 2.13% improvement in nDCG@10, 13.11% increase in HR@10, and a 17.43% reduction in RMSE on the Google Local Reviews dataset. These findings highlight the transformative potential of ESE-DT in advancing personalized recommendation systems by integrating emotion and sentiment analysis with the powerful capabilities of the Decision Transformer.

Atherosclerotic plaque may rupture without warning, causing drastic clinical events such as heart attack and stroke. Methods to predict plaque vulnerability change is lacking in the literature. Three predictive methods were introduced to identify best morphological and mechanical predictors for plaque morphological vulnerability change. Patient follow-up in vivo intravascular ultrasound (IVUS) coronary plaque data were acquired from nine individuals to construct fluid-structure interaction models and obtain plaque biomechanical conditions. Morphological plaque vulnerability index (MPVI) was defined using features related to plaque rupture and used as a substitute measure for plaque vulnerability since clinical plaque rupture data is not available for this study. All combinations of ten biomechanical and morphological factors (see Table 1) were tested by generalized linear mixed model (GLMM), support vector machine (SVM) and random forest (RF) using data from 374 IVUS slices at baseline to identify the optimal predictor for plaque vulnerability change: MPVI (MPVI follow-up – MPVI baseline ). Prediction results showed that baseline MPVI was the best single risk factor for predicting MPVI using GLMM and RF with prediction accuracy 82.09% and 78.53%, respectively, while baseline plaque area (PA) was the best one for SVM. The combination of mean wall thickness, lumen area, PA, critical plaque wall stress, and MPVI was the best overall predictor using RF with the highest prediction accuracy 91.47%, 5.91% improvement over that using GLMM. These results demonstrated that combining morphological and mechanical risk factors could lead to better prediction accuracy. The outcome of plaque vulnerability prediction would provide another dimension for decision making in clinical patient screening, diagnosis and intervention.