Hydrological modelling using machine and deep learning models across multiple case studies

Background/Objectives: Cytochrome P450 3A4 (CYP3A4) metabolizes approximately 30–50% of clinically used drugs; thus, accurate prediction of CYP3A4 inhibition is essential for early assessment of drug–drug interaction (DDI) risk and toxicity. This study evaluated an integrated artificial intelligence framework for predicting CYP3A4 inhibition (%) using a large, curated chemical dataset. Methods: A dataset of 23,713 compounds was compiled from the Korea Chemical Bank and multiple commercial and public databases. Vector-based machine learning (ML) models (LightGBM, XGBoost, CatBoost, and a weighted ML ensemble) and graph neural network (GNN) models (O-GNN with contrastive learning and manifold mixup (O-GNN + CL + Mixup), D-MPNN, GINE, and GATv2) were evaluated. Manifold mixup was applied during GNN training, and SMILES enumeration-based test-time augmentation was used at inference. The best-performing ML and GNN models were integrated using a weighted ensemble strategy. Model interpretability was examined using SHAP analysis for ML models and occlusion sensitivity analysis for O-GNN + CL + Mixup. Results: The weighted ML ensemble achieved the best performance among ML models (RMSE = 19.1031, Pearson correlation coefficient (PCC) = 0.7566); the O-GNN + CL + Mixup model performed the best among GNN models (RMSE = 20.1002, PCC = 0.7265). The hybrid model achieved improved predictive accuracy (RMSE = 19.0784, PCC = 0.7570). External validation on 100 newly generated experimental data points confirmed generalizability (Custom Metric = 0.8035). Conclusions: This study demonstrated that integrating ML and GNN models with data augmentation strategies improves the robustness and interpretability of CYP3A4 inhibition prediction and established a practical framework for metabolic screening and DDI risk assessment.

ObjectiveThis retrospective, case-control study with internal validation evaluates the performance of machine learning (ML) and deep learning (DL) models in classifying pediatric patients at risk for anxiety disorders using structured electronic health records (EHRs) and area-based measures of health (ABMH). The aim is to enable proactive care by monitoring potential anxiety onset across developmental stages.MethodsWe trained a series of ML models (Logistic Regression, Decision Tree, Random Forest, K-Nearest Neighbors, XGBoost) and DL models (LSTM, GRU, RETAIN, Dipole) using structured EHR data from 30-day windows prior to diagnosis. Two datasets were used per age group: one with structured EHR data only, and another including both EHR and ABMH data. ML models were trained using short-term cross-sectional features, while DL models leveraged full longitudinal patient histories. Performance was assessed using AUROC, AUPRC, PPV, NPV, F1 score, and accuracy. Due to differences in input scope, model performance reflects both algorithmic and temporal design differences and is not intended as a direct comparison between ML and DL.ResultsML models offered strong baseline performance, with XGBoost achieving AUROC scores of 0.817 (EHR) and 0.816 (EHR+ABMH) for 8-year-olds. Adding ABMH features did not significantly improve performance. DL models, particularly RETAIN and Dipole, achieved the highest AUROC values (e.g., Dipole: 0.853 with EHR, 0.857 with EHR+ABMH for 8-year-olds), outperforming other DL and ML models within their respective design constraints.ConclusionBoth ML and DL models successfully identified likely anxiety onset using structured EHR data. DL models using longitudinal data achieved the highest performance, while XGBoost provided a robust ML baseline. The minimal impact of ABMH features highlights integration challenges, and performance variation across ages emphasizes the need for age-stratified modeling approaches.

Failure to predict stroke promptly may lead to delayed treatment, causing severe consequences like permanent neurological damage or death. Early detection using deep learning (DL) and machine learning (ML) models can enhance patient outcomes and mitigate the long-term effects of strokes. The aim of this study is to compare these models, exploring their efficacy in predicting stroke. This study analyzed a dataset comprising 663 records from patients hospitalized at Hazrat Rasool Akram Hospital in Tehran, Iran, including 401 healthy individuals and 262 stroke patients. A total of eight established ML (SVM, XGB, KNN, RF) and DL (DNN, FNN, LSTM, CNN) models were utilized to predict stroke. Techniques such as 10-fold cross-validation and hyperparameter tuning were implemented to prevent overfitting. The study also focused on interpretability through Shapley Additive Explanations (SHAP). The evaluation of model’s performance was based on accuracy, specificity, sensitivity, F1-score, and ROC curve metrics. Among DL models, LSTM showed superior sensitivity at 96.15%, while FNN exhibited better specificity (96.0%), accuracy (96.0%), F1-score (95.0%), and ROC (98.0%) among DL models. For ML models, RF displayed higher sensitivity (99.9%), accuracy (99.0%), specificity (100%), F1-score (99.0%), and ROC (99.9%). Overall, RF outperformed all models, while DL models surpassed ML models in most metrics except for RF. DL models (CNN, LSTM, DNN, FNN) achieved sensitivities from 93.0 to 96.15%, specificities from 80.0 to 96.0%, accuracies from 92.0 to 96.0%, F1-scores from 87.34 to 95.0%, and ROC scores from 95.0 to 98.0%. In contrast, ML models (KNN, XGB, SVM) showed sensitivities between 29.0% and 94.0%, specificities between 89.47% and 96.0%, accuracies between 71.0% and 95.0%, F1-scores between 44.0% and 95.0%, and ROC scores between 64.0% and 95.0%. This study demonstrates the efficacy of DL and ML models in predicting stroke, with the RF models outperforming all others in key metrics. While DL models generally surpassed ML models, RF’s exceptional performance highlights the potential of combining these technologies for early stroke detection, significantly improving patient outcomes by preventing severe consequences like permanent neurological damage or death.

This study presents a comprehensive comparative analysis of Machine Learning (ML) and Deep Learning (DL) models for predicting Wind Turbine (WT) power output based on environmental variables such as temperature, humidity, wind speed, and wind direction. Along with Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), and Convolutional Neural Network (CNN), the following ML models were looked at: Linear Regression (LR), Support Vector Regressor (SVR), Random Forest (RF), Extra Trees (ET), Adaptive Boosting (AdaBoost), Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM). Using a dataset of 40,000 observations, the models were assessed based on R-squared, Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). ET achieved the highest performance among ML models, with an R-squared value of 0.7231 and a RMSE of 0.1512. Among DL models, ANN demonstrated the best performance, achieving an R-squared value of 0.7248 and a RMSE of 0.1516. The results show that DL models, especially ANN, did slightly better than the best ML models. This means that they are better at modeling non-linear dependencies in multivariate data. Preprocessing techniques, including feature scaling and parameter tuning, improved model performance by enhancing data consistency and optimizing hyperparameters. When compared to previous benchmarks, the performance of both ANN and ET demonstrates significant predictive accuracy gains in WT power output forecasting. This study’s novelty lies in directly comparing a diverse range of ML and DL algorithms while highlighting the potential of advanced computational approaches for renewable energy optimization.

Human-induced global warming, primarily attributed to the rise in atmospheric CO2, poses a substantial risk to the survival of humanity. While most research focuses on predicting annual CO2 emissions, which are crucial for setting long-term emission mitigation targets, the precise prediction of daily CO2 emissions is equally vital for setting short-term targets. This study examines the performance of 14 models in predicting daily CO2 emissions data from 1/1/2022 to 30/9/2023 across the top four polluting regions (China, India, the USA, and the EU27&UK). The 14 models used in the study include four statistical models (ARMA, ARIMA, SARMA, and SARIMA), three machine learning models (support vector machine (SVM), random forest (RF), and gradient boosting (GB)), and seven deep learning models (artificial neural network (ANN), recurrent neural network variations such as gated recurrent unit (GRU), long short-term memory (LSTM), bidirectional-LSTM (BILSTM), and three hybrid combinations of CNN-RNN). Performance evaluation employs four metrics (R2, MAE, RMSE, and MAPE). The results show that the machine learning (ML) and deep learning (DL) models, with higher R2 (0.714–0.932) and lower RMSE (0.480–0.247) values, respectively, outperformed the statistical model, which had R2 (− 0.060–0.719) and RMSE (1.695–0.537) values, in predicting daily CO2 emissions across all four regions. The performance of the ML and DL models was further enhanced by differencing, a technique that improves accuracy by ensuring stationarity and creating additional features and patterns from which the model can learn. Additionally, applying ensemble techniques such as bagging and voting improved the performance of the ML models by approximately 9.6%, whereas hybrid combinations of CNN-RNN enhanced the performance of the RNN models. In summary, the performance of both the ML and DL models was relatively similar. However, due to the high computational requirements associated with DL models, the recommended models for daily CO2 emission prediction are ML models using the ensemble technique of voting and bagging. This model can assist in accurately forecasting daily emissions, aiding authorities in setting targets for CO2 emission reduction.

BackgroundThe proportion of Canadian youth seeking mental health support from an emergency department (ED) has risen in recent years. As EDs typically address urgent mental health crises, revisiting an ED may represent unmet mental health needs. Accurate ED revisit prediction could aid early intervention and ensure efficient healthcare resource allocation. We examine the potential increased accuracy and performance of graph neural network (GNN) machine learning models compared to recurrent neural network (RNN), and baseline conventional machine learning and regression models for predicting ED revisit in electronic health record (EHR) data.MethodsThis study used EHR data for children and youth aged 4–17 seeking services at McMaster Children’s Hospital’s Child and Youth Mental Health Program outpatient service to develop and evaluate GNN and RNN models to predict whether a child/youth with an ED visit had an ED revisit within 30 days. GNN and RNN models were developed and compared against conventional baseline models. Model performance for GNN, RNN, XGBoost, decision tree and logistic regression models was evaluated using F1 scores.ResultsThe GNN model outperformed the RNN model by an F1-score increase of 0.0511 and the best performing conventional machine learning model by an F1-score increase of 0.0470. Precision, recall, receiver operating characteristic (ROC) curves, and positive and negative predictive values showed that the GNN model performed the best, and the RNN model performed similarly to the XGBoost model. Performance increases were most noticeable for recall and negative predictive value than for precision and positive predictive value.ConclusionsThis study demonstrates the improved accuracy and potential utility of GNN models in predicting ED revisits among children and youth, although model performance may not be sufficient for clinical implementation. Given the improvements in recall and negative predictive value, GNN models should be further explored to develop algorithms that can inform clinical decision-making in ways that facilitate targeted interventions, optimize resource allocation, and improve outcomes for children and youth.

The need for intubation in methanol-poisoned patients, if not predicted in time, can lead to irreparable complications and even death. Artificial intelligence (AI) techniques like machine learning (ML) and deep learning (DL) greatly aid in accurately predicting intubation needs for methanol-poisoned patients. So, our study aims to assess Explainable Artificial Intelligence (XAI) for predicting intubation necessity in methanol-poisoned patients, comparing deep learning and machine learning models. This study analyzed a dataset of 897 patient records from Loghman Hakim Hospital in Tehran, Iran, encompassing cases of methanol poisoning, including those requiring intubation (202 cases) and those not requiring it (695 cases). Eight established ML (SVM, XGB, DT, RF) and DL (DNN, FNN, LSTM, CNN) models were used. Techniques such as tenfold cross-validation and hyperparameter tuning were applied to prevent overfitting. The study also focused on interpretability through SHAP and LIME methods. Model performance was evaluated based on accuracy, specificity, sensitivity, F1-score, and ROC curve metrics. Among DL models, LSTM showed superior performance in accuracy (94.0%), sensitivity (99.0%), specificity (94.0%), and F1-score (97.0%). CNN led in ROC with 78.0%. For ML models, RF excelled in accuracy (97.0%) and specificity (100%), followed by XGB with sensitivity (99.37%), F1-score (98.27%), and ROC (96.08%). Overall, RF and XGB outperformed other models, with accuracy (97.0%) and specificity (100%) for RF, and sensitivity (99.37%), F1-score (98.27%), and ROC (96.08%) for XGB. ML models surpassed DL models across all metrics, with accuracies from 93.0% to 97.0% for DL and 93.0% to 99.0% for ML. Sensitivities ranged from 98.0% to 99.37% for DL and 93.0% to 99.0% for ML. DL models achieved specificities from 78.0% to 94.0%, while ML models ranged from 93.0% to 100%. F1-scores for DL were between 93.0% and 97.0%, and for ML between 96.0% and 98.27%. DL models scored ROC between 68.0% and 78.0%, while ML models ranged from 84.0% to 96.08%. Key features for predicting intubation necessity include GCS at admission, ICU admission, age, longer folic acid therapy duration, elevated BUN and AST levels, VBG_HCO3 at initial record, and hemodialysis presence. This study as the showcases XAI's effectiveness in predicting intubation necessity in methanol-poisoned patients. ML models, particularly RF and XGB, outperform DL counterparts, underscoring their potential for clinical decision-making.

Anaplastic lymphoma kinase (ALK), a tyrosine kinase receptor, is identified as a crucial target in the progression of anticancer therapeutics for non-small cell lung cancer. This study has executed a Graph Neural Network (GNN) model and compared it with three machine learning (ML) models based on fingerprints for rapid anticancer bioactivity prediction. ALK inhibitors with IC50 values were extracted from the REAXYS database. Following preprocessing, these inhibitors constituted a dataset of 1664 molecules. Subsequently, GNN and ML models were constructed on a training set. The generalizability of these models was assessed by internal and external validation procedures. The graph neural network model yielded promising results, with an average precision of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.879\pm 0.041$</tex> and an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F1$</tex> score of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.804\pm 0.049$</tex> in cross-validation. In external validation, the model achieved an average precision of 0.938 and an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F1$</tex> score of 0.863, surpassing the results of the ML models. Therefore, we can infer that the predictive model developed using the GNN is apt for the problem at hand and can be utilized to predict the biological activity of novel ALK inhibitors.

Data augmentation using SMOTE technique: Application for prediction of burst pressure of hydrocarbons pipeline using supervised machine learning models

The digital era has transformed the way businesses interact with their customers, with online platforms serving as crucial touchpoints for user engagement. Understanding customer behavior in this context is paramount for enhancing user experience, optimizing marketing strategies, and driving business growth. This study aims to explore the likelihood of customers making purchases based on their clickstream data by employing both machine learning and deep learning techniques. This research uses a machine learning model Random Forest (RF), Gradient Boosting Decision Trees (GBDT), Extreme Gradient Boosting (XGBOOST) and deep learning model Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM) to predict whether customers will purchase the items using 33,040,175 records in the file of the click and 1,177,769 records in the buys file from real e-commerce customers. The results show that both machine learning and deep learning can accurately forecast the purchasing behavior of customers with an accuracy of around 72 to 75 percent. For the machine learning model, attains the highest prediction accuracy when using a sliding window of 6 days. For the deep learning model, the LSTM model with 50 layers shows the highest prediction of customers willingness to purchase an item. Compared with previous studies, the three machine learning models narrow the range of days, give more accurate predictions, and also improve the model. Both RNN and LSTM show similar accuracy for customer behavior. The current research has asserted that both machine learning and deep learning models give profound results on whether customers will purchase a product, and there is not a significant difference between machine learning and deep learning in this classification topic.

This paper discusses the compressive strength prediction for self-compacting concrete (SCC) by a host of machine learning (ML) and deep learning (DL) models is discussed in this research work. Random Forest (RF), Keras Regressor (KR), Extremely Randomized Trees (ERT), Extreme Gradient Boosting (XGB), Gradient Boosting (GB), Light Gradient Boosting Machine (LGBM), and Category Boosting (CB) are some of the many ensemble methods until now. In addition, the ability of several models to predict the compressive strength of SCC was examined with generalized additive models like Gradient Boosting Regressor and Neural Networks based on Keras. Twenty papers constituted the dataset, which was divided into three subsets for validation, testing, and training. The principal input parameters utilized in model building are superplasticizers, cement, water, fine aggregates, coarse aggregates, and mineral admixtures. To check the accuracy of each model developed, some performance indicators were chosen, like R², RMSE, MAE, and MAPE, which measure how accurately a model predicts compressive strength. The best predictive accuracy was found for the models under test in GB with R² = 5.12, MSE = 26.23, and MAE = 4.13, whereas Keras Regressor also performed very well with R² = 0.6948, RMSE = 0.0832, and MAE = 0.0569. These results thus establish that the GB and KR models can prove to be good resources for predictive efficiency in determining the compressive strength of SCC, exhibiting great potential for machine learning and deep learning methodologies applied to concrete materials.

In this study, four machine learning (ML) models (gradient boost decision tree (GBDT), light gradient boosting machine (LightGBM), categorical boosting (CatBoost) and extreme gradient boosting (XGBoost)) are used to perform seasonal forecasts for non-monsoonal winter precipitation over the Eurasian continent (30-60°N, 30-105°E) (NWPE). The seasonal forecast results from a traditional linear regression (LR) model and two dynamic models are compared. The ML and LR models are trained using the data for the period of 1979-2010, and then, these empirical models are used to perform the seasonal forecast of NWPE for 2011-2018. Our results show that the four ML models have reasonable seasonal forecast skills for the NWPE and clearly outperform the LR model. The ML models and the dynamic models have skillful forecasts for the NWPE over different regions. The ensemble means of the forecasts including the ML models and dynamic models show higher forecast skill for the NWEP than the ensemble mean of the dynamic-only models. The forecast skill of the ML models mainly benefits from a skillful forecast of the third empirical orthogonal function (EOF) mode (EOF3) of the NWPE, which has a good and consistent prediction among the ML models. Our results also illustrate that the sea ice over the Arctic in the previous autumn is the most important predictor in the ML models in forecasting the NWPE. This study suggests that ML models may be useful tools to help improve seasonal forecasts of the NWPE.

Improving GALDIT-based groundwater vulnerability predictive mapping using coupled resampling algorithms and machine learning models

The thesis evaluates the machine learning (ML) and deep learning (DL) approaches’ performance in accurately detecting glaucoma based on optical coherence tomography tabular data and images from individuals of different ethnicities. While numerous studies have employed ML and DL techniques for glaucoma identification, their performance has not been evaluated across diverse ethnic groups. In addition, a DL approach utilizing the Swin Transformer architecture trained on the thickness map images of the retinal nerve fiber layer (RNFL) was also evaluated. This Swin transformer DL model demonstrated an AUC of 0.97 in the internal testing dataset (Asian) and 0.88 in the external testing dataset (Caucasian). However, like the ML classifiers trained on measured data, the DL approach which was trained on raw thickness map images also exhibited poor reproducibility across different datasets. To address these issues, a cross-sectional study design was employed to investigate both ML and DL’s model performance in glaucoma detection using OCT data from individuals of different ethnicities. The study included 514 Asian participants, consisting of 257 with glaucoma and 257 controls, to develop ML and DL classifiers. The trained classifiers were subsequently evaluated on two separate participant groups comprising 356 Asians and 138 Caucasians. Two machine learning classifiers were created using the two types of RNFL thickness, one using the original values extracted from OCT machines (measured RNFL), and the other generated from the compensation model. The compensation model is a multivariate regression trained on normal individuals. It corrects the 12-clock RNFL thicknesses for multiple demographic and anatomical parameters. Additionally, a deep learning model was developed using the Swin Transformer architecture based on the measured RNFL thickness map images from OCT. Explainable artificial intelligence techniques (CAM and SHAP) were utilized to better interpret the results. Performance metrics such as the area under the receiver operating characteristic curve (AUC), accuracy and sensitivity were employed to examine the effectiveness of different glaucoma detection models. Both machine learning (AUC = 0.96) and deep learning (AUC = 0.97) models demonstrated superior performance compared to the raw measured data (baseline, AUC = 0.93), in the internal testing dataset (Asian). However, in the external testing dataset (Caucasian), ML models utilizing the compensated data (AUC = 0.93) exhibited significantly better performance compared to ML models using the original measured data (AUC = 0.83) and the baseline (AUC = 0.82). Furthermore, the ML and DL models trained on measured data exhibited inadequate generalization ability across different ethnicities, whereas the ML model using the compensated data maintained its performance in the external testing dataset. These findings caution against the indiscriminate application of ML and DL models to patient cohorts of different ethnicities. They also suggest that incorporating the compensation model into the development of ML models may enhance their performance in glaucoma detection across diverse ethnicities. Overall, the study highlights the importance of accounting for anatomical variations across different ethnic groups when developing machine-learning models for glaucoma detection using OCT data.

In recent years, numerous Machine Learning (ML) models, including Deep Learning (DL) and classic ML models, have been developed to detect software vulnerabilities. However, there is a notable lack of comprehensive and systematic surveys that summarize, classify, and analyze the applications of these ML models in software vulnerability detection. This absence may lead to critical research areas being overlooked or under-represented, resulting in a skewed understanding of the current state of the art in software vulnerability detection. To close this gap, we propose a comprehensive and systematic literature review that characterizes the different properties of ML-based software vulnerability detection systems using six major Research Questions (RQs). Using a custom web scraper, our systematic approach involves extracting a set of studies from four widely used online digital libraries: ACM Digital Library, IEEE Xplore, ScienceDirect, and Google Scholar. We manually analyzed the extracted studies to filter out irrelevant work unrelated to software vulnerability detection, followed by creating taxonomies and addressing RQs. Our analysis indicates a significant upward trend in applying ML techniques for software vulnerability detection over the past few years, with many studies published in recent years. Prominent conference venues include the International Conference on Software Engineering (ICSE), the International Symposium on Software Reliability Engineering (ISSRE), the Mining Software Repositories (MSR) conference, and the ACM International Conference on the Foundations of Software Engineering (FSE), whereas Information and Software Technology (IST), Computers &amp; Security (C&amp;S), and Journal of Systems and Software (JSS) are the leading journal venues. Our results reveal that 39.1% of the subject studies use hybrid sources, whereas 37.6% of the subject studies utilize benchmark data for software vulnerability detection. Code-based data are the most commonly used data type among subject studies, with source code being the predominant subtype. Graph-based and token-based input representations are the most popular techniques, accounting for 57.2% and 24.6% of the subject studies, respectively. Among the input embedding techniques, graph embedding and token vector embedding are the most frequently used techniques, accounting for 32.6% and 29.7% of the subject studies. Additionally, 88.4% of the subject studies use DL models, with recurrent neural networks and graph neural networks being the most popular subcategories, whereas only 7.2% use classic ML models. Among the vulnerability types covered by the subject studies, CWE-119, CWE-20, and CWE-190 are the most frequent ones. In terms of tools used for software vulnerability detection, Keras with TensorFlow backend and PyTorch libraries are the most frequently used model-building tools, accounting for 42 studies for each. In addition, Joern is the most popular tool used for code representation, accounting for 24 studies. Finally, we summarize the challenges and future directions in the context of software vulnerability detection, providing valuable insights for researchers and practitioners in the field.