A Deep Learning Approach Based on Interpretable Feature Importance for Predicting Sports Results

Accurately predicting the clinical trajectory of patients with implantable cardioverter-defibrillators (ICDs) is critical for guiding their care and management. Machine learning (ML) methods surpass traditional statistical approaches by addressing complex data patterns and variability, providing more precise and personalized risk estimates. This retrospective study included patients from four major hospitals in China. Data from three hospitals were used for training and internal tests, while data from the remaining hospital were used for external tests. Six ML models were developed and validated. Model discrimination was measured using the area under the receiver operating characteristic curve (AUROC). Kaplan-Meier survival curves were generated by stratifying patients into high-risk and low-risk groups based on the optimal model's predictions. Interpretation analysis was performed to rank the importance of predictive features. A total of 3175 patients were studied. The multilayer perceptron (MLP) model demonstrated superior predictive accuracy, with the AUROC of 0.70 and 0.72 in internal and external test sets, respectively, outperforming other models. Kaplan-Meier curves show distinct survival trends over time between high-risk and low-risk groups, with stratification determined by the MLP model using a Youden's index cut-off value of 0.3443 (p < 0.001). Among the seven key predictors identified, glomerular filtration rate (GFR) was the most influential factor. The MLP model effectively predicted 3-year survival for ICD or cardiac resynchronization therapy defibrillator (CRT-D) patients and accurately stratified them into distinct risk groups. The integration of MLP and SHapley Additive exPlanations (SHAP) provided explicit explanations for individualized risk predictions, facilitated clinical decision-making, and supported the optimization of treatment strategies. ClinicalTrials.gov identifier: NCT05396313.

BackgroundIndividuals with chronic diseases are at higher risk of sarcopenia, and precise prediction is essential for its prevention. This study aims to develop a risk scoring model using longitudinal data to predict the probability of sarcopenia in this population over next 3–5 years, thereby enabling early warning and intervention.MethodsUsing data from a nationwide survey initiated in 2011, we selected patient data records from wave 1 (2011–2012) and follow-up data from wave 3 (2015–2016) as the study cohort. Retrospective data collection included demographic information, health conditions, and biochemical markers. After excluding records with missing values, a total of 2891 adults with chronic conditions were enrolled. Sarcopenia was assessed based on the Asian Working Group for Sarcopenia (AWGS) 2019 guidelines. A generalized linear mixed model (GLMM) with random effects and diverse machine learning models were utilized to explore feature contributions to sarcopenia risk. The Recursive Feature Elimination (RFE) algorithm was employed to optimize the full Multilayer Perceptron (MLP) model and develop an online application tool.ResultsAmong total population, 580 (20.1%) individuals were diagnosed with sarcopenia in wave 1 (2011–2012), and 638 (22.1%) were diagnosed in wave 3 (2015–2016), while 2165 (74.9%) individuals were not diagnosed with sarcopenia across the study period. MLP model, performed better than other three classic machine learning models, demonstrated a ROC AUC of 0.912, a PR AUC of 0.401, a sensitivity of 0.875, a specificity of 0.844, a Kappa value of 0.376, and an F1 score of 0.44. According to MLP model-based SHapley Additive exPlanations (SHAP) scoring, weight, age, BMI, height, total cholesterol, PEF, and gender were identified as the most important features of chronic disease individuals for sarcopenia. Using the RFE algorithm, we selected six key variables—weight, age, BMI, height, total cholesterol, and gender—achieving an ROC AUC of about 0.9 for the online application tool.ConclusionWe developed an MLP machine learning model that incorporates only six easily accessible variables, enabling the prediction of sarcopenia risk in individuals with chronic diseases. Additionally, we created a practical online application tool to assist in decision-making and streamline clinical assessments.

Classifying demonstration format and presenter identity in imitative learning task: EEG-based explainable machine learning.

The objective of this study is to develop artificial neural network (ANN) models, including multilayer perceptron (MLP) and Kohonen self-organizing feature map (KSOFM), for spatial disaggregation of areal rainfall in the Wi-stream catchment, an International Hydrological Program (IHP) representative catchment, in South Korea. A three-layer MLP model, using three training algorithms, was used to estimate areal rainfall. The Levenberg–Marquardt training algorithm was found to be more sensitive to the number of hidden nodes than were the conjugate gradient and quickprop training algorithms using the MLP model. Results showed that the networks structures of 11-5-1 (conjugate gradient and quickprop) and 11-3-1 (Levenberg-Marquardt) were the best for estimating areal rainfall using the MLP model. The networks structures of 1-5-11 (conjugate gradient and quickprop) and 1-3-11 (Levenberg–Marquardt), which are the inverse networks for estimating areal rainfall using the best MLP model, were identified for spatial disaggregation of areal rainfall using the MLP model. The KSOFM model was compared with the MLP model for spatial disaggregation of areal rainfall. The MLP and KSOFM models could disaggregate areal rainfall into individual point rainfall with spatial concepts.

Both statistical and neural methods have been proposed in the literature to predict healthcare expenditures. However, less attention has been given to comparing predictions from both these methods as well as ensemble approaches in the healthcare domain. The primary objective of this paper was to evaluate different statistical, neural, and ensemble techniques in their ability to predict patients' weekly average expenditures on certain pain medications. Two statistical models, persistence (baseline) and autoregressive integrated moving average (ARIMA), a multilayer perceptron (MLP) model, a long short-term memory (LSTM) model, and an ensemble model combining predictions of the ARIMA, MLP, and LSTM models were calibrated to predict the expenditures on two different pain medications. In the MLP and LSTM models, we compared the influence of shuffling of training data and dropout of certain nodes in MLPs and nodes and recurrent connections in LSTMs in layers during training. Results revealed that the ensemble model outperformed the persistence, ARIMA, MLP, and LSTM models across both pain medications. In general, not shuffling the training data and adding the dropout helped the MLP models and shuffling the training data and not adding the dropout helped the LSTM models across both medications. We highlight the implications of using statistical, neural, and ensemble methods for time-series forecasting of outcomes in the healthcare domain.

Early detection of sepsis is crucial for timely intervention and improved patient outcomes. Traditional diagnostic methods often fail to identify sepsis in its initial stages, leading to delays in treatment. This study investigates the potential of machine learning techniques, particularly multilayer perceptron (MLP) models, for early sepsis prediction using physiological data from intensive care unit (ICU) patients. A large dataset comprising approximately 40,000 ICU patient records from two hospital systems was analyzed. Various machine learning algorithms were implemented and compared, with a focus on optimizing an MLP model for early sepsis detection. The MLP model was trained on physiological data collected up to six hours before the clinical manifestation of sepsis symptoms. The findings demonstrated that the MLP model outperformed traditional methods, accurately predicting sepsis onset up to six hours before clinical symptoms became apparent. The MLP model exhibited significant improvements in accuracy compared to conventional models used for sepsis diagnosis. The ability to predict sepsis development at an early stage holds immense clinical value. Early detection facilitated by the MLP model can potentially lead to prompt administration of appropriate treatments, thereby improving patient outcomes and reducing mortality rates associated with sepsis. Moreover, early identification of sepsis cases can optimize resource allocation and improve operational efficiency in critical care environments. This study highlights the efficacy of MLP models in leveraging physiological data for early sepsis prediction. The proposed approach offers a promising solution for enhancing sepsis management protocols, ultimately contributing to improved patient care and resource utilization in intensive care settings

The utilization of steel slag for CO2 sequestration is an effective way to reduce carbon emissions. The reactivity of steel slag in CO2 sequestration depends mainly on material and process parameters. However, there are many puzzles in regard to practical applications due to the different evaluations of process parameters and the lack of investigation of material parameters. In this study, 318 samples were collected to investigate the interactive influence of 12 factors on the carbonation reactivity of steel slag by machine learning with SHapley Additive exPlanations (SHAP). Multilayer perceptron (MLP), random forest, and support vector regression models were built to predict the slurry-phase CO2 sequestration of steel slag. The MLP model performed well in terms of prediction ability and generalization with comprehensive interpretability. The SHAP results showed that the impact of the process parameters was greater than that of the material parameters. Interestingly, the iron ore phase of steel slag was revealed to have a positive effect on steel slag carbonation by SHAP analysis. Combined with previous literature, the carbonation mechanism of steel slag was proposed. Quantitative analysis based on SHAP indicated that steel slag had good carbonation reactivity when the mass fractions of "CaO + MgO", "SiO2 + Al2O3", "Fe2O3", and "MnO" varied from 50-55%, 10-15%, 30-35%, and <5%, respectively.

Uveal melanoma (UM) patients face a significant risk of distant metastasis, closely tied to a poor prognosis. Despite this, there is a dearth of research utilizing big data to predict UM distant metastasis. This study leveraged machine learning methods on the Surveillance, Epidemiology, and End Results (SEER) database to forecast the risk probability of distant metastasis. Therefore, the information on UM patients from the SEER database (2000-2020) was split into a 7:3 ratio training set and an internal test set based on distant metastasis presence. Univariate and multivariate logistic regression analyses assessed distant metastasis risk factors. Six machine learning methods constructed a predictive model post-feature variable selection. The model evaluation identified the multilayer perceptron (MLP) as optimal. Shapley additive explanations (SHAP) interpreted the chosen model. A web-based calculator personalized risk probabilities for UM patients. The results show that nine feature variables contributed to the machine learning model. The MLP model demonstrated superior predictive accuracy (Precision = 0.788; ROC AUC = 0.876; PR AUC = 0.788). Grade recode, age, primary site, time from diagnosis to treatment initiation, and total number of malignant tumors were identified as distant metastasis risk factors. Diagnostic method, laterality, rural-urban continuum code, and radiation recode emerged as protective factors. The developed web calculator utilizes the MLP model for personalized risk assessments. In conclusion, the MLP machine learning model emerges as the optimal tool for predicting distant metastasis in UM patients. This model facilitates personalized risk assessments, empowering early and tailored treatment strategies.

Anthropogenic activities, species invasions, and ecological factors are driving rapid changes in rangeland ecosystems. For ensuring the richness and sustainability of plant species habitats, there is a pressing need for a reliable prediction model that can accurately forecast and map species distribution under varying ecological conditions. We aimed to compare the performance of three widely used machine learning methods: Multilayer perceptron (MLP), radial basis function (RBF), and support vector machine (SVM) in predicting the distribution ofFestuca ovinain mountainous protected rangelands. We conducted our investigation by analyzing the distribution ofF. ovinain 305 randomly selected plant sample plots. In each plot, we recorded 10 ecological variables. Three machine learning models were developed to predict the likelihood ofF. ovinadistribution. Our results demonstrated that the RBF model had a higher number of misclassifications (11 samples) compared to MLP and SVM models (10 samples), indicating that MLP and SVM models were more accurate for distribution modeling. Additionally, MLP showed a higher R‐squared value (0.87) compared to SVM (0.85), suggesting that MLP was the most accurate model for restoring degraded lands. Hence, we developed theFestuca ovinaDistribution Model (FODM) using the MLP model. Sensitivity analyses revealed that soil texture, soil depth, electrical conductivity (EC), pH, and vegetation density significantly influencedF. ovinadistribution, with respective sensitivity coefficients of 0.48, 0.47, 0.45, 0.41, and 0.41. Based on the finalized FODM, we designed an Environmental Decision Support System (EDSS) tool to assist rangeland managers in mappingF. ovinadistribution. By applying the EDSS tool, we demonstrated its practicality in using the FODM for effective decision‐making and land management. The EDSS tool serves as a valuable resource for rangeland managers, enabling them to make informed decisions regardingF. ovinarestoration and effectively use the predictive capabilities of the FODM in real‐world applications.

Evapotranspiration (ET) is of paramount importance due to its crucial role in the water cycle, moving water from land to the atmosphere. This process is critical for sustaining atmospheric rivers and guiding water management operations. Process-based hydrological modeling is commonly used to predict ET in various ecosystems. However, while plant growth dynamics are better understood in temperate regions, the accuracy of ET predictions in tropical areas remains limited. This reduced accuracy is primarily due to challenges in simulating the Leaf Area Index (LAI), the intensity of mass and energy exchanges, and the prevalence of energy-limited conditions.In this study, we explore the potential of data-driven models to estimate LAI in the tropics and improve ET predictions. We implemented a Multilayer Perceptron (MLP) model trained using climatological variables from ERA-5 and CHIRPS as inputs. The model's performance was evaluated using LAI values from MODIS at the Cesar River Watershed in northern Colombia, South America.Comparisons between the selected MLP model and SWAT reveal an improvement over the default LAI simulated by the latter. Particularly, SWAT underestimates foliage growth and fails to capture the bimodal behavior observed in the study area. The MLP model, tested at the watershed and Hydrologic Response Unit (HRU) scales, demonstrated promising results. The performance of the proposed MLP model was evaluated using shuffled and sequential schemes, achieving validation Nash-Sutcliffe efficiencies between 0.5 and 0.99 at the tested scales. In addition, the results show that the MLP model is especially sensitive to the seasonal component of relative humidity. By leveraging remote sensing data, data-driven models become a potential tool to simulate remote sensed LAI with greater accuracy. This potential of the MLP model to significantly improve LAI and ET predictions can enhance hydrologic models' reliability, especially under shifting environmental conditions, and offers an enhanced outlook for better simulating multiple water compartments in the tropics.

ABSTRACTThe traditional structural finite element seismic response analysis process is often complex and time consuming. This study introduces a novel prediction methodology utilizing machine learning (ML) models to facilitate the rapid analysis of the seismic fragility of high‐rise shear wall structures. Three high‐rise shear wall structural models with varying story heights were designed in compliance with current Chinese codes. Ground motion records were selected as inputs based on the target response spectrum, and a sample database for ML was constructed using the incremental dynamic analysis (IDA) method. Key ground motion intensity and structural parameters were chosen as the characteristic inputs, with the maximum interstory drift angle (θmax) and maximum floor acceleration serving as the output parameters. Structural responses were predicted using both the multilayer perceptron (MLP) and random forest (RF) methods. The MLP model was further analyzed using SHapley Additive exPlanations (SHAP) to explore the contributions of each characteristic parameter to the structural response. The seismic fragility of the high‐rise shear wall structures was subsequently assessed. Results indicate that the RF‐based method provides higher accuracy in seismic fragility analysis compared with the MLP algorithm. Notably, peak ground acceleration (PGA) emerged as the most significant parameter impacting structural response. The proposed methodology effectively enables rapid seismic fragility analysis of high‐rise shear wall structures.

Basal stem rot disease (BSR) in oil palm plants is caused by the Ganoderma boninense (G. boninense) fungus. BSR is a major disease that affects oil palm plantations in Malaysia and Indonesia. As of now, the only available sustaining measure is to prolong the life of oil palm trees since there has been no effective treatment for the BSR disease. This project used an ALOS PALSAR-2 image with dual polarization, Horizontal transmit and Horizontal receive (HH) and Horizontal transmit and Vertical receive (HV). The aims of this study were to (1) identify the potential backscatter variables; and (2) examine the performance of machine learning (ML) classifiers (Multilayer Perceptron (MLP) and Random Forest (RF) to classify oil palm trees that are non-infected and infected by G. boninense. The sample size consisted of 55 uninfected trees and 37 infected trees. We used the imbalance data approach (Synthetic Minority Over-Sampling Technique (SMOTE) in these classifications due to the differing sample sizes. The result showed backscatter variable HV had a higher correct classification for the G. boninense non-infected and infected oil palm trees for both classifiers; the MLP classifier model had a robust success rate, which correctly classified 100% for non-infected and 91.30% for infected G. boninense, and RF had a robust success rate, which correctly classified 94.11% for non-infected and 91.30% for infected G. boninense. In terms of model performance using the most significant variables, HV, the MLP model had a balanced accuracy (BCR) of 95.65% compared to 92.70% for the RF model. Comparison between the MLP model and RF model for the receiver operating characteristics (ROC) curve region, (AUC) gave a value of 0.92 and 0.95, respectively, for the MLP and RF models. Therefore, it can be concluded by using only the HV polarization, that both the MLP and RF can be used to predict BSR disease with a relatively high accuracy.

Influent flow rate is a crucial parameter closely related to the plant-wide control of wastewater treatment plants (WWTPs). In this study, a random forest (RF) model and a multi-layer perceptron (MLP) model are developed for hourly influent flow rate prediction at a confidential WWTP in Canada. Both models perform well on predicting influent flow rate one-step ahead. The coefficient of determination (R2) values of MLP and RF for the testing data set are 0.927 and 0.925, respectively. Furthermore, the multi-step ahead prediction accuracy of the proposed models is discussed. To improve the multi-step ahead prediction accuracy of the RF model, time-tag information is transformed to numerical values and then fed into the RF model as input. The R2 values of the RF model for the testing data set with and without time-tag information are 0.334 and 0.811, respectively. The results show that the RF model’s performance for multi- step ahead prediction is heavily affected by the time-tag information. Including time-tag information as input could dramatically improve the multi-step ahead prediction accuracy. In this study, the RF model shows more robust performance than the MLP model on solving short-term wastewater influent prediction problems. Influent flow rate is a crucial parameter closely related to the plant-wide control of wastewater treatment plants (WWTPs). In this study, a random forest (RF) model and a multi-layer perceptron (MLP) model are developed for hourly influent flow rate prediction at a confidential WWTP in Canada. Both models perform well on predicting influent flow rate one-step ahead. The coefficient of determination (R2) values of MLP and RF for the testing data set are 0.927 and 0.925, respectively. Furthermore, the multi-step ahead prediction accuracy of the proposed models is discussed. To improve the multi-step ahead prediction accuracy of the RF model, time-tag information is transformed to numerical values and then fed into the RF model as input. The R2 values of the RF model for the testing data set with and without time-tag information are 0.334 and 0.811, respectively. The results show that the RF model’s performance for multi-step ahead prediction is heavily affected by the time-tag information. Including time-tag information as input could dramatically improve the multi-step ahead prediction accuracy. In this study, the RF model shows more robust performance than the MLP model on solving short-term wastewater influent prediction problems.

In this paper, we have attempted to build multilayer perceptron (MLP) and support vector machine (SVM) models for predicting survivability of the oral cancer patients who visit the ENT OPD. MLP and SVM have been applied in the past by few researchers for prediction of oral cancer using the genetic database. However, the database used for current research has the attributes like clinical symptoms, history of addiction, diagnosis, investigations, treatments and follow-up details which is gathered from presentations and review graphs related to oral malignancy from ENT and head and neck department. The MLP and SVM models are compared on the basis of various estimation criteria to identify the most effective model. Experimental result shows that accuracy of classification of SVM model is 73.56 %, whereas MLP model is 70.05 %; specificity of SVM model is 73.53 %, whereas MLP model is 65.36 %; and sensitivity of MLP model is 77.00 %, whereas SVM model is 73.56 %. SVM displays better results in terms of true negative, false negative, geometric mean of sensitivity and specificity, positive predictive value, geometric mean of positive predictive value and negative predictive value, precision, F-measure, area under receiver operating characteristics curve and lift and gain chart. Hence, it may be concluded that SVM is a most favourable model for predicting survival rate of oral cancer patients.

Postoperative ileus (POI) continues to be a major cause of morbidity following colorectal surgery. Despite best efforts, the incidence of POI in colorectal surgery remains high (~30%). This study aimed to investigate machine learning techniques to identify risk factors for POI in colorectal surgery patients, to help guide further preventative strategies. A TRIPOD-guideline-compliant retrospective study was conducted for major colorectal surgery patients at a single tertial care centre (2018-2022). The primary outcome was the occurrence of POI, defined as not achieving GI-2 (outcome measure of time to first stool and tolerance of oral diet) by day four. Multivariate logistic regression, decision trees, radial basis function and multilayer perceptron (MLP) models were trained using a random allocation of patients to training/testing data sets (80/20). The area under the receiver operating characteristic (AUROC) curves were used to evaluate model performance. Of 504 colorectal surgery patients, 183 (36%) experienced POI. Multivariate logistic regression, decision trees, radial basis function and MLP models returned an AUROC of 0.722, 0.706, 0.712 and 0.800, respectively. The MLP model had the highest sensitivity and specificity values. In addition to well-known risk factors for POI, such as postoperative hypokalaemia, surgical approach, and opioid use, the MLP model identified sarcopenia (ranked 4/30) as a potentially modifiable risk factor for POI. MLP outperformed other models in predicting POI. Machine learning can provide valuable insights into the importance and ranking of specific predictive variables for POI. Further research into the predictive value of preoperative sarcopenia for POI is required.