Diverse Machine Learning Research Articles

Machine Learning-Based Prediction for Incident Hypertension Based on Regular Health Checkup Data: Derivation and Validation in 2 Independent Nationwide Cohorts in South Korea and Japan.

Worldwide, cardiovascular diseases are the primary cause of death, with hypertension as a key contributor. In 2019, cardiovascular diseases led to 17.9 million deaths, predicted to reach 23 million by 2030. This study presents a new method to predict hypertension using demographic data, using 6 machine learning models for enhanced reliability and applicability. The goal is to harness artificial intelligence for early and accurate hypertension diagnosis across diverse populations. Data from 2 national cohort studies, National Health Insurance Service-National Sample Cohort (South Korea, n=244,814), conducted between 2002 and 2013 were used to train and test machine learning models designed to anticipate incident hypertension within 5 years of a health checkup involving those aged ≥20 years, and Japanese Medical Data Center cohort (Japan, n=1,296,649) were used for extra validation. An ensemble from 6 diverse machine learning models was used to identify the 5 most salient features contributing to hypertension by presenting a feature importance analysis to confirm the contribution of each future. The Adaptive Boosting and logistic regression ensemble showed superior balanced accuracy (0.812, sensitivity 0.806, specificity 0.818, and area under the receiver operating characteristic curve 0.901). The 5 key hypertension indicators were age, diastolic blood pressure, BMI, systolic blood pressure, and fasting blood glucose. The Japanese Medical Data Center cohort dataset (extra validation set) corroborated these findings (balanced accuracy 0.741 and area under the receiver operating characteristic curve 0.824). The ensemble model was integrated into a public web portal for predicting hypertension onset based on health checkup data. Comparative evaluation of our machine learning models against classical statistical models across 2 distinct studies emphasized the former's enhanced stability, generalizability, and reproducibility in predicting hypertension onset.

Integration of transcriptomics and machine learning for insights into breast cancer: exploring lipid metabolism and immune interactions

BackgroundBreast cancer (BRCA) represents a substantial global health challenge marked by inadequate early detection rates. The complex interplay between the tumor immune microenvironment and fatty acid metabolism in BRCA requires further investigation to elucidate the specific role of lipid metabolism in this disease.MethodsWe systematically integrated nine machine learning algorithms into 184 unique combinations to develop a consensus model for lipid metabolism-related prognostic genes (LMPGS). Additionally, transcriptomics analysis provided a comprehensive understanding of this prognostic signature. Using the ESTIMATE method, we evaluated immune infiltration among different risk subgroups and assessed their responsiveness to immunotherapy. Tailored treatments were screened for specific risk subgroups. Finally, we verified the expression of key genes through in vitro experiments.ResultsWe identified 259 differentially expressed genes (DEGs) related to lipid metabolism through analysis of the cancer genome atlas program (TCGA) database. Subsequently, via univariate Cox regression analysis and C-index analysis, we developed an optimal machine learning algorithm to construct a 21-gene LMPGS model. We used optimal cutoff values to divide the lipid metabolism prognostic gene scores into two groups according to high and low scores. Our study revealed distinct biological functions and mutation landscapes between high-scoring and low-scoring patients. The low-scoring group presented a greater immune score, whereas the high-scoring group presented enhanced responses to both immunotherapy and chemotherapy drugs. Single-cell analysis highlighted significant upregulation of CPNE3 in epithelial cells. Moreover, by employing molecular docking, we identified niclosamide as a potential targeted therapeutic drug. Finally, our experiments demonstrated high expression of MTMR9 and CPNE3 in BRCA and their significant correlation with prognosis.ConclusionBy employing bioinformatics and diverse machine learning algorithms, we successfully identified genes associated with lipid metabolism in BRCA and uncovered potential therapeutic agents, thereby offering novel insights into the mechanisms and treatment strategies for BRCA.

AI fusion of multisource data identifies key features of vitiligo

Vitiligo is a skin disorder that is associated with a decreased risk of skin cancer, but it can lead to increased susceptibility to sunburn, psychological distress, and disruptions in daily life, consists of two primary subtypes: segmental and nonsegmental vitiligo, each with distinct underlying mechanisms. However, the reliable identification of diagnostic markers and the ability to differentiate between these subtypes have remained elusive challenges. This study aims to pioneer predictive algorithms for vitiligo diagnosis, harnessing the capabilities of AI (Artificial Intelligence) to amalgamate multisource data and uncover essential features for distinguishing vitiligo subtypes.An ensemble algorithm was thoughtfully developed for vitiligo diagnosis, utilizing a spectrum of machine learning techniques to evaluate the likelihood of vitiligo, whether segmental or nonsegmental. Diverse machine learning methodologies were applied to distinguish between healthy individuals and vitiligo patients, as well as to differentiate segmental from nonsegmental vitiligo. The ensemble algorithm achieved a remarkable AUC (Area Under the Curve) of 0.99 and an accuracy of 0.98 for diagnosing vitiligo. Furthermore, in predicting the development of segmental or nonsegmental vitiligo, the model exhibited an AUC of 0.79 and an accuracy of 0.73. Key parameters for vitiligo identification encompassed factors such as age, FBC (full blood count)-neutrophils, FBC-lymphocytes, LKF(liver and kidney function)-direct bilirubin, LKF-total bilirubin, and LKF-total protein levels. In contrast, vital indicators for monitoring the progression of segmental and nonsegmental vitiligo included FBC-B lymphocyte count, FBC-NK (Natural Killer) cell count, and LKF-alkaline phosphatase levels. This retrospective study underscores the potential of AI-driven analysis in identifying significant risk factors for vitiligo and predicting its subtypes at an early stage. These findings offer great promise for the development of effective diagnostic tools and the implementation of personalized treatment approaches in managing this challenging skin disorder.

Integrated approach of machine learning, Mendelian randomization and experimental validation for biomarker discovery in diabetic nephropathy.

To identify potential biomarkers and explore the mechanisms underlying diabetic nephropathy (DN) by integrating machine learning, Mendelian randomization (MR) and experimental validation. Microarray and RNA-sequencing datasets (GSE47184, GSE96804, GSE104948, GSE104954, GSE142025 and GSE175759) were obtained from the Gene Expression Omnibus database. Differential expression analysis identified the differentially expressed genes (DEGs) between patients with DN and controls. Diverse machine learning algorithms, including least absolute shrinkage and selection operator, support vector machine-recursive feature elimination, and random forest, were used to enhance gene selection accuracy and predictive power. We integrated summary-level data from genome-wide association studies on DN with expression quantitative trait loci data to identify genes with potential causal relationships to DN. The predictive performance of the biomarker gene was validated using receiver operating characteristic (ROC) curves. Gene set enrichment and correlation analyses were conducted to investigate potential mechanisms. Finally, the biomarker gene was validated using quantitative real-time polymerase chain reaction in clinical samples from patients with DN and controls. Based on identified 314 DEGs, seven characteristic genes with high predictive performance were identified using three integrated machine learning algorithms. MR analysis revealed 219 genes with significant causal effects on DN, ultimately identifying one co-expressed gene, carbonic anhydrase II (CA2), as a key biomarker for DN. The ROC curves demonstrated the excellent predictive performance of CA2, with area under the curve values consistently above 0.878 across all datasets. Additionally, our analysis indicated a significant association between CA2 and infiltrating immune cells in DN, providing potential mechanistic insights. This biomarker was validated using clinical samples, confirming the reliability of our findings in clinical practice. By integrating machine learning, MR and experimental validation, we successfully identified and validated CA2 as a promising biomarker for DN with excellent predictive performance. The biomarker may play a role in the pathogenesis and progression of DN via immune-related pathways. These findings provide important insights into the molecular mechanisms underlying DN and may inform the development of personalized treatment strategies for this disease.

A novel triage framework for emergency department based on machine learning paradigm

AbstractEmergency departments, crucial in managing patient emergencies, are often challenged by overcrowding and diagnostic errors during triage—the process that assesses the urgency of patients' conditions. Traditional triage systems, heavily dependent on human judgement, are prone to errors like under‐triage, where severe conditions are missed, delaying treatment, and over‐triage, where less severe conditions are overly prioritized, causing unnecessary resource use and morbidity. This thesis presents a novel multi‐model machine‐learning framework designed to enhance triage accuracy by evaluating multiple medical conditions concurrently, including chronic illnesses like heart disease. The framework employs various machine‐learning techniques—such as logistic regression, support vector machines, random forests, deep neural networks, and decision trees—to analyze different medical conditions in two comprehensive phases. In the first phase, patients' conditions are categorized, and a multi‐tiered analysis using multiple classifiers refines the assessment by considering probabilistic outcomes from each classifier. The second phase synthesizes these insights into a unified and precise triage decision, distinguishing between patients who require critical care and those needing less urgent hospitalization. This integration of diverse machine learning models allows for a fused and precise triage decision, overcoming traditional triage systems' limitations that usually focus on single conditions and rely on isolated model predictions. A hybrid feature selection method is also utilized to identify critical predictors, enhancing the decision‐making process. The framework is validated through a specially curated dataset that simulates multiple triage scenarios, evaluated by medical experts for efficacy. Comparative analysis with traditional triage methods demonstrates significant improvements in decision accuracy, as evidenced by higher area under curve (AUC) values—0.95 for critical care and 0.90 for hospitalization. The implementation of this data‐driven approach substantially reduces human error, boosts operational efficiency, and aids medical staff in making rapid, informed decisions. This research represents a significant advancement in emergency medical care, illustrating the benefits of integrating sophisticated machine learning techniques to improve patient outcomes and resource management.

Relationship matters: Using machine learning methods to predict the mental health severity of Chinese college freshmen during the pandemic period

BackgroundPandemics act as stressors and may lead to frequent mental health disorders. College student, especially freshmen, are particularly susceptible to experiencing intense mental stress reactions during a pandemic. We aimed to identify stable and intervenable variables including academic, relationship and economic factors, and focused on their impact on mental health severity during the pandemic period. MethodsWe innovatively combined diverse machine learning methods, including XGBoost, SHAP, and K-means clustering, to predict the mental health severity of college freshmen. A total of 3281 college freshmen participated in the research. Discriminant analyses were performed on groups of participants with depression (PHQ-9), anxiety (GAD-7). All characteristic variables were selected based on their importance and interventionability. Further analyses were conducted with selected features to determine the optimal variable combination. ResultsXGBoost analysis revealed that relationship factors exhibited the highest predictive capacity for mental health severity among college freshmen (SHAPFamily Relationship = 0.373; SHAPPeer Support = 0.236). The impact of academic factors on college freshmen's mental health severity depended on their intricate interplay with relationship factors, resulting in complex interactive effects. These effects were heterogeneous among different subgroups. ConclusionsThe proposed machine learning approach utilizing XGBoost, SHAP and K-means clustering methods provides a valuable tool to gain insights into the relative contributions of academic, relationship and economic factors to Chinese college freshmen's mental health severity during the COVID-19 pandemic. The result guide the development of targeted intervention measures tailored to meet specific requirements within each subgroup.

Machine learning-based discovery of UPP1 as a key oncogene in tumorigenesis and immune escape in gliomas.

Gliomas are the most common and aggressive type of primary brain tumor, with a poor prognosis despite current treatment approaches. Understanding the molecular mechanisms underlying glioma development and progression is critical for improving therapies and patient outcomes. The current study comprehensively analyzed large-scale single-cell RNA sequencing and bulk RNA sequencing of glioma samples. By utilizing a series of advanced computational methods, this integrative approach identified the gene UPP1 (Uridine Phosphorylase 1) as a novel driver of glioma tumorigenesis and immune evasion. High levels of UPP1 were linked to poor survival rates in patients. Functional experiments demonstrated that UPP1 promotes tumor cell proliferation and invasion and suppresses anti-tumor immune responses. Moreover, UPP1 was found to be an effective predictor of mutation patterns, drug response, immunotherapy effectiveness, and immune characteristics. These findings highlight the power of combining diverse machine learning methods to identify valuable clinical markers involved in glioma pathogenesis. Identifying UPP1 as a tumor growth and immune escape driver may be a promising therapeutic target for this devastating disease.

Regulating Modality Utilization within Multimodal Fusion Networks.

Multimodal fusion networks play a pivotal role in leveraging diverse sources of information for enhanced machine learning applications in aerial imagery. However, current approaches often suffer from a bias towards certain modalities, diminishing the potential benefits of multimodal data. This paper addresses this issue by proposing a novel modality utilization-based training method for multimodal fusion networks. The method aims to guide the network's utilization on its input modalities, ensuring a balanced integration of complementary information streams, effectively mitigating the overutilization of dominant modalities. The method is validated on multimodal aerial imagery classification and image segmentation tasks, effectively maintaining modality utilization within ±10% of the user-defined target utilization and demonstrating the versatility and efficacy of the proposed method across various applications. Furthermore, the study explores the robustness of the fusion networks against noise in input modalities, a crucial aspect in real-world scenarios. The method showcases better noise robustness by maintaining performance amidst environmental changes affecting different aerial imagery sensing modalities. The network trained with 75.0% EO utilization achieves significantly better accuracy (81.4%) in noisy conditions (noise variance = 0.12) compared to traditional training methods with 99.59% EO utilization (73.7%). Additionally, it maintains an average accuracy of 85.0% across different noise levels, outperforming the traditional method's average accuracy of 81.9%. Overall, the proposed approach presents a significant step towards harnessing the full potential of multimodal data fusion in diverse machine learning applications such as robotics, healthcare, satellite imagery, and defense applications.

Human lung cancer classification and comprehensive analysis using different machine learning techniques.

Lung cancer is the most common causes of death among all cancer-related diseases. A lung scan examination of the patient is the primary diagnostic technique. This scan analysis pertains to an MRI, CT, or X-ray. The automated classification of lung cancer is difficult due to the involvement of multiple steps in imaging patients' lungs. In this manuscript, human lung cancer classification and comprehensive analysis using different machine learning techniques is proposed. Initially, the input images are gathered using lung cancer dataset. The proposed method processes these images using image-processing techniques, and further machine learning techniques are utilized for categorization. Seven different classifiers including the k-nearest neighbors (KNN), support vector machine (SVM), decision tree (DT), multinomial naive Bayes (MNB), stochastic gradient descent (SGD), random forest (RF), and multi-layer perceptron (MLP) classifier are used, which classifies the lung cancer as malignant and benign. The performance of the proposed approach is examined using performances metrics, like positive predictive value, accuracy, sensitivity, and f-score are evaluated. Among them, the performance of the MLP classifier provides 25.34%, 45.39%, 15.39%, 41.28%, 22.17%, and 12.12% higher accuracy than other KNN, SVM, DT, MNB, SGD, and RF respectively. RESEARCH HIGHLIGHTS: Lung cancer is a leading cause of cancer-related death. Imaging (MRI, CT, and X-ray) aids diagnosis. Automated classification of lung cancer faces challenges due to complex imaging steps. This study proposes human lung cancer classification using diverse machine learning techniques. Input images from lung cancer dataset undergo image processing and machine learning. Classifiers like k-nearest neighbors, support vector machine, decision tree, multinomial naive Bayes, stochastic gradient descent, random forest, and multi-layer perceptron (MLP) classify cancer types; MLP excels in accuracy.

Precision Medicine With Data-driven Approaches: A Framework For Clinical Translation

Precision medicine-based approaches differentiate themselves by taking into consideration subpopulation variability (e.g. genetic variations, age, gender, race, addictions). To date, traditional models, such as Trial-and-Error Dosing, Empirical Treatment Guidelines, Statistical and Actuarial Models, Pathophysiological Models, and Clinical Judgment and Experience, have been generalized in healthcare fields. However, more comprehensive and innovative technologies are required besides these conventional modelings, which are subject to various limitations such as low efficiency and incapability of processing complex biological systems. Here we review diverse machine learning (ML) algorithms integrated with big data and omics and its applications in various aspects of precision medicine. ML is the branch of artificial intelligence (AI), which has been rapidly developed and highlighted as a promising method to decrease diagnostic errors and aid clinicians with decision-making in recent decades. We focused on applications of ML models such as support vector machine (SVM), K Nearest Neighbor (KNN) random forest (RF), convolutional neural networks (CNNs) and deep learning in drug toxicity prediction, cardiovascular diseases, neurodegenerative diseases, and cancer therapies within precision medicine and specific benefits and challenges of each. This review provides insights on the wider utilization in clinical environments by recognizing current advantages that are expected to expand the scope of AI-driven methods and issues that need to be addressed for further studies.

Multiple Machine Learning Models in House Price Prediction: Performance Evaluation and Comparison

House pricing is one of the most critical concerns for people. Accurately predicting house prices is essential for participants and investors in the real estate market. This study delves into the realm of predictive modeling for house prices using a multimodal approach in machine learning. The research methodology spans various stages including data collection, preprocessing, feature engineering, and model selection. Based on a dataset sourced from Kaggle, housing data from Iowa comprising 79 pertinent features influencing property prices is utilized. Performance evaluation is conducted using the root mean square error (RMSE), comparing the efficacy of linear regression models such as Lasso regression, Kernel Ridge regression, and ElasticNet regression against tree-based models including XGBoost, Gradient Boosting, and LightGBM. Additionally, the study implements two stacked ensemble models to identify the most optimal predictive model for house prices. Experimental procedures involve training and evaluating these models on specific datasets to gauge their predictive accuracy. Findings and discussions underscore the comparative analysis of diverse machine learning models within the context of house price prediction. Moreover, the study employs the Gradient Boosting method in machine learning to pinpoint the ten most influential features impacting house prices. Recommendations for future research are provided to enhance the precision and robustness of house price prediction models. This research contributes to the advancement of predictive modeling in the real estate domain, offering insights into the efficacy of various machine learning techniques for forecasting property prices.

Leveraging multiple data types for improved compound-kinase bioactivity prediction

Machine learning provides efficient ways to map compound-kinase interactions. However, diverse bioactivity data types, including single-dose and multi-dose-response assay results, present challenges. Traditional models utilize only multi-dose data, overlooking information contained in single-dose measurements. Here, we propose a machine learning methodology for compound-kinase activity prediction that leverages both single-dose and dose-response data. We demonstrate that our two-stage approach yields accurate activity predictions and significantly improves model performance compared to training solely on dose-response labels. This superior performance is consistent across five diverse machine learning methods. Using the best performing model, we carried out extensive experimental profiling on a total of 347 selected compound-kinase pairs, achieving a high hit rate of 40% and a negative predictive value of 78%. We show that these rates can be improved further by incorporating model uncertainty estimates into the compound selection process. By integrating multiple activity data types, we demonstrate that our approach holds promise for facilitating the development of training activity datasets in a more efficient and cost-effective way.

Quantifying seasonal variations in pollution sources with machine learning-enhanced positive matrix factorization

As the pace of industrialization and urbanization accelerates, water quality management faces increasing challenges, with traditional methods for pollutant source apportionment often proving inadequate in handling complex environmental data. This study enhances the precision and reliability of pollutant source identification by integrating Positive Matrix Factorization (PMF) models with diverse machine learning techniques. Utilizing data from 17 water quality monitoring stations along the Wuding River from 2017 to 2021, we employed Random Forest (RF), Support Vector Machine (SVM), Elastic Net (EN), and Extreme Gradient Boosting (XGBoost) models to predict the Water Quality Index (WQI) during dry and wet seasons. Results indicate that the RF model exhibited optimal performance in the dry season (R2 = 0.93), while the SVM was superior in the wet season (R2 = 0.94). SHAP (SHapley Additive exPlanations) value analysis identified CODMn and NH3-N as significant influencers on WQI in the dry season, whereas COD, BOD, and TP gained prominence during the wet season. SHAP values reveal the contribution of each feature to the model output, thereby enhancing the model’s transparency and interpretability. Additionally, feature importance identified by machine learning was utilized as weights to optimize the contribution rates predicted by the PMF model. The optimised model was able to identify the contribution of domestic and farm effluent discharges more accurately in the dry season, with a significant increase in the percentage of identification from 19.4 % to 45.4 %, and an increase in the percentage of contribution from agricultural non-point sources and domestic effluent in the rainy season. This research offers a novel perspective on the characteristics of river water pollution and holds significant implications for formulating data-driven environmental management strategies.

COMPARISON OF MACHINE LEARNING TECHNIQUES FOR CLASSIFICATION OF DISTRIBUTED DENIAL OF SERVICE ATTACKS BASED ON FEATURE ENGINEERING IN SDN-BASED NETWORKS

Distributed Denial-of-Service (DDoS) attacks present a noteworthy cybersecurity hazard to software-defined networks (SDNs). This investigation presents an approach that depends on feature engineering and machine learning to discern DDoS attacks in SDNs. Initially, the dataset acquired from Kaggle goes through cleansing and normalization procedures, and the optimal subset of features is determined by employing the Correlation-based Feature Selection (CFS) algorithm. Subsequently, the optimal subset of features is trained and evaluated utilizing diverse Machine Learning algorithms, specifically Random Forest (RF), Decision Tree, Adaptive Boosting (AdaBoost), K-Nearest Neighbor (k-NN), Gradient Boosting, Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Categorical Boosting (CatBoost). The outcomes demonstrate that XGBoost outperforms the other algorithms in various performance metrics (e.g., accuracy, precision, recall, F1, and AUC values). Furthermore, a comparative analysis was carried out among various models and algorithms, revealing that the technique proposed by the researchers yielded the most favourable outcomes and effectively detected and identified DDoS attacks in SDN. Consequently, this investigation provides a novel perspective and resolution for SDN security.

Enhancing solar photovoltaic energy production prediction using diverse machine learning models tuned with the chimp optimization algorithm

Solar photovoltaic (PV) systems, integral for sustainable energy, face challenges in forecasting due to the unpredictable nature of environmental factors influencing energy output. This study explores five distinct machine learning (ML) models which are built and compared to predict energy production based on four independent weather variables: wind speed, relative humidity, ambient temperature, and solar irradiation. The evaluated models include multiple linear regression (MLR), decision tree regression (DTR), random forest regression (RFR), support vector regression (SVR), and multi-layer perceptron (MLP). These models were hyperparameter tuned using chimp optimization algorithm (ChOA) for a performance appraisal. The models are subsequently validated on the data from a 264 kWp PV system, installed at the Applied Science University (ASU) in Amman, Jordan. Of all 5 models, MLP shows best root mean square error (RMSE), with the corresponding value of 0.503, followed by mean absolute error (MAE) of 0.397 and a coefficient of determination (R2) value of 0.99 in predicting energy from the observed environmental parameters. Finally, the process highlights the fact that fine-tuning of ML models for improved prediction accuracy in energy production domain still involves the use of advanced optimization techniques like ChOA, compared with other widely used optimization algorithms from the literature.

Extreme flash flood susceptibility mapping using a novel PCA-based model stacking approach

This study introduces an efficient methodology for model stacking, incorporating six diverse machine learning and statistical models alongside principal component analysis (PCA). The approach is applied for the flash flood susceptibility mapping within the Karkheh Basin in Iran. The selected models include random forest (RF), boosted regression trees (BRT), support vector machine (SVM), artificial neural networks (ANN), generalized additive model (GAM), and the least absolute shrinkage and selection operator (Lasso), with RF also serving as the meta-model for the stacking. The results revealed significant correlations among the predictions of the individual models, which could potentially impact the meta-model’s efficacy. To address this, PCA was applied to the model predictions to generate de-correlated components as inputs for the meta-model, thereby enhancing prediction accuracy and robustness. Evaluation based on the area under the receiver operating characteristic (AUROC) curve demonstrated that the GAM outperformed all other individual models with the highest accuracy score of 0.924. In contrast, the RF and ANN models had the lowest accuracy, both registering at 0.872. However, the performance disparity across models was minimal. Notably, the PCA-based stacking approach (0.936) surpassed both traditional model stacking (0.912) and the performances of all individual models, advocating for its use in enhancing predictive accuracy. These findings endorse the PCA-stacking method over conventional stacking techniques. Nonetheless, further research across varied applications is warranted to generalize its efficacy.

Intelligent diagnosis of Kawasaki disease from real-world data using interpretable machine learning models

ObjectiveThis study aimed to leverage real-world electronic medical record data to develop interpretable machine learning models for diagnosis of Kawasaki disease while also exploring and prioritizing the significant risk factors. MethodsA comprehensive study was conducted on 4087 pediatric patients at the Children’s Hospital of Chongqing, China. The study collected demographic data, physical examination results, and laboratory findings. Statistical analyses were performed using IBM SPSS Statistics, Version 26.0. The optimal feature subset was used to develop intelligent diagnostic prediction models based on the Light Gradient Boosting Machine, Explainable Boosting Machine (EBM), Gradient Boosting Classifier (GBC), Fast Interpretable Greedy-Tree Sums, Decision Tree, AdaBoost Classifier, and Logistic Regression. Model performance was evaluated in three dimensions: discriminative ability via receiver operating characteristic curves, calibration accuracy using calibration curves, and interpretability through SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations). ResultsIn this study, Kawasaki disease was diagnosed in 2971 participants. Analysis was conducted on 31 indicators, including red blood cell distribution width and erythrocyte sedimentation rate. The EBM model demonstrated superior performance relative to other models, with an area under the curve of 0.97, second only to the GBC model. Furthermore, the EBM model exhibited the highest calibration accuracy and maintained its interpretability without relying on external analytical tools such as SHAP and LIME, thus reducing interpretation biases. Platelet distribution width, total protein, and erythrocyte sedimentation rate were identified by the model as significant predictors for the diagnosis of Kawasaki disease. ConclusionThis study used diverse machine learning models for early diagnosis of Kawasaki disease. The findings demonstrated that interpretable models such as EBM outperformed traditional machine learning models in terms of both interpretability and performance. Ensuring consistency between predictive models and clinical evidence is crucial for the successful integration of artificial intelligence into real-world clinical practice.

An Investigation of Machine Learning Applications in the Financial Fraud Detection

Financial fraud presents itself in various forms, and it often involves intricate financial transaction networks, making it challenging to detect the perpetrators and identify the characteristics of the fraud. In recent years, machine learning has gained widespread application within the financial sector. Therefore, various financial fraud detection models have been developed based on diverse machine learning methodologies. In the method part, this paper provides an overview of the machine learning process and then discusses the application of machine learning models in various financial fraud scenarios. Financial fraud in the insurance field has been further refined into automobile and medical insurance fraud. In automobile insurance fraud detection, some studies applied implicit Naive Bayes Model to analyze observable features and estimate hidden variables, and some studies used resampler to solve data imbalance and adopted 7 kinds of machine learning models for analysis. In health insurance fraud detection, many studies train medical data with multiple models, including AdaBoost, Logistic Regression and Support Vector Machine. In credit card fraud detection, studies use five algorithms, including random forest and decision tree et al., and some construct a model combining Decision Tree (DT) and Logistic Regression (LR). In bank fraud detection, some studies introduce Value-at-Risk to combine it with machine learning algorithms, and some studies propose a decentralized model training method based on federated learning. In the discussion section, this paper addresses the current limitations of the research, such as poor interpretability, uneven distribution of data sets, and issues related to customer privacy, and proposes corresponding solutions.

Cancer classification using RNA sequencing gene expression data based on Game Shapley local search embedded binary social ski-driver optimization algorithms

Cancer remains a significant health concern due to its high mortality rates. Recent decades have witnessed substantial progress in cancer research, driven by advancements in high throughput sequencing technology and the application of diverse machine learning (ML) methods, particularly in the analysis of gene expression data. However, the proliferation of high-dimensional datasets, such as RNA-seq data, underscores the need for more robust ML techniques capable of efficiently handling large volumes of data to enable accurate treatment decisions. This paper introduces a novel hybrid feature selection (FS) algorithm, termed Game kernel SHapley Additive exPlanations (kSHAP), which combines with binary Social Ski Driver (bSSD), Adaptive Beta Hill Climbing (ABHC) and Late Acceptance Hill Climbing (LAHC) algorithms. The study comprehensively investigates three novel FS algorithms—kSHAP-bSSD, kSHAP-ABHC, and kSHAP-LAHC for cancer classification tasks using RNA sequencing (RNA-seq) datasets. An experiment conducted on five well-established RNA-seq cancer datasets: Lung Adenocarcinoma (LUAD), Stomach Adenocarcinoma (STAD), Breast Invasive Carcinoma (BRCA), lung squamous cell carcinoma (LUSC) and uterine corpus endometrial carcinoma (UCEC). The objective is to enhance cancer classification accuracy, robustness, and scalability using RNA-seq datasets. Additionally, the study evaluates six classifiers—Autoencoder (AE), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naive Bayes (NB), Neural Network (NN), and Random Forest (RF) with AE consistently out performing others. Evaluation metrics include accuracy, recall, precision, box plot, F1-score, radar plot, confusion matrix, ROC and statistical analysis. Our approach is compared against recent state-of-the-art FS algorithms, showing improvements in gene selection and classification accuracy. The kSHAP-bSSD demonstrates superior performance across all datasets compared to traditional methods, achieving an accuracy rate of 99.9% in LUAD and exhibiting higher accuracy rates and robustness in STAD, BRCA, LUSC, and UCEC datasets. Assessment across multiple metrics affirms the superiority of kSHAP-bSSD and kSHAP-ABHC combinations, underscoring their effectiveness in cancer classification tasks.

The robust SmartEnsembleNet: A game changer in finger knuckle biometrics

SmartEnsembleNet represents a major leap forward in the domain of biometric identification, focusing on the relatively underexplored area of finger knuckle biometrics. This research enhances the effectiveness and uniqueness of biometric systems by leveraging the distinct and secure features of finger knuckle patterns. As a complement to existing biometric techniques like fingerprints and facial recognition, it adds an additional layer of security and reliability. Addressing the challenges faced by finger knuckle identification, such as image acquisition, environmental effects, and aging, SmartEnsembleNet introduces a pioneering multi-layered ensemble learning strategy that combines eight diverse machine learning models. These models, trained via a 5-fold cross-validation on datasets from IIT Delhi and PolyU, demonstrate significant improvements in accuracy, precision, recall, and F1-score. A distinctive feature of SmartEnsembleNet is its innovative meta-model, streamlining the creation of a meta-dataset. This dataset is created by applying bilinear interpolation to the predicted probabilities from the top-performing model, and it is further refined by subtracting these interpolated probabilities from the validation data. This iterative process generates ‘smart data’, which serves as the basis for training the ultimate ‘smart model’. Logistic regression is chosen for its outstanding performance. This method, applied uniformly to both training and testing phases, considerably enhances classification accuracy in finger knuckle biometrics, establishing new benchmarks. SmartEnsembleNet’s performance undergoes a detailed evaluation using the Friedman test and Nemenyi post-hoc analysis, guaranteeing a selection process grounded in statistical rigor. With its resilience to noise and adaptability under various conditions, SmartEnsembleNet not only addresses existing gaps in finger knuckle biometric research but also pioneers future advancements, providing a versatile and secure solution impactful for both academia and high-security scenarios.