Decision trees: from efficient prediction to responsible AI

Blockeel, Hendrik; Devos, Laurens; Frénay‬, Benoît; Nanfack, Géraldin; Nijssen, Siegfried

doi:10.3389/frai.2023.1124553

Cited by 21 publications

(13 citation statements)

References 148 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Models using XGBoost (Extreme Gradient Boosting), decision trees and K-nearest neighbor are attractive options. [40][41][42] Agnostic approaches such as unsupervised ML and deep learning (DL) may also be used, and these may identify novel signals in the ECG associated with LVH.…”

Section: Discussionmentioning

confidence: 99%

Diagnostic and prognostic value of ECG-predicted hypertension-mediated left ventricular hypertrophy using machine learning

Naderi,

Ramírez,

van Duijvenboden

et al. 2024

Preprint

View full text Add to dashboard Cite

Background Four hypertension-mediated left ventricular hypertrophy (LVH) phenotypes have been reported using cardiac magnetic resonance (CMR): normal LV, LV remodeling, eccentric and concentric LVH, with varying prognostic implications. The electrocardiogram (ECG) is routinely used to detect LVH, however its capacity to differentiate between LVH phenotypes is unknown. This study aimed to classify hypertension-mediated LVH from the ECG using machine learning (ML) and test for associations of ECG-predicted phenotypes with incident cardiovascular outcomes. Methods ECG biomarkers were extracted from the 12-lead ECG of 20,439 hypertensives in UK Biobank (UKB). Classification models integrating ECG and clinical variables were built using logistic regression, support vector machine (SVM) and random forest. The models were trained in 80% of participants, and the remaining 20% formed the test set. External validation was sought in 877 hypertensives from Study of Health in Pomerania (SHIP). In the UKB test set, we tested for associations between ECG-predicted LVH phenotypes and incident major adverse cardiovascular events (MACE) and heart failure. Results Among UKB participants 19,408 had normal LV, 758 LV remodeling, 181 eccentric and 92 concentric LVH. Classification performance of the three models was comparable, with SVM having a slightly superior performance (accuracy 0.79 ,sensitivity 0.59, specificity 0.87, AUC 0.69) and similar results observed in SHIP. There was superior prediction of eccentric LVH in both cohorts. In the UKB test set, ECG-predicted eccentric LVH was associated with heart failure (HR 3.42, CI 1.06-9.86). Conclusions ECG-based ML classifiers represent a potentially accessible screening strategy for early detection of hypertension-mediated LVH phenotypes.

show abstract

Section: Discussionmentioning

confidence: 99%

Diagnostic and prognostic value of ECG-predicted hypertension-mediated left ventricular hypertrophy using machine learning

Naderi,

Ramírez,

van Duijvenboden

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Decision tree is a basic classification and regression method. This study uses a classification decision tree, which has a tree-shaped structure and consists of two parts: nodes and directed edges [ 29 ].After repeated training and optimization, the parameters of the decision tree C5.0 model constructed in this study.…”

Section: Methodsmentioning

confidence: 99%

Prognostic value of plasma microRNAs for non-small cell lung cancer based on data mining models

Yan,

Kang,

Liu

et al. 2024

BMC Cancer

View full text Add to dashboard Cite

Background As biomarkers, microRNAs (miRNAs) are closely associated with the occurrence, progression, and prognosis of non-small cell lung cancer (NSCLC). However, the prognostic predictive value of miRNAs in NSCLC has rarely been explored. In this study, the value in prognosis prediction of NSCLC was mined based on data mining models using clinical data and plasma miRNAs biomarkers. Methods A total of 69 patients were included in this prospective cohort study. After informed consent, they filled out questionnaires and had their peripheral blood collected. The expressions of plasma miRNAs were examined by quantitative polymerase chain reaction (qPCR). The Whitney U test was used to analyze non-normally distributed data. Kaplan-Meier was used to plot the survival curve, the log-rank test was used to compare with the overall survival curve, and the Cox proportional hazards model was used to screen the factors related to the prognosis of lung cancer. Data mining techniques were utilized to predict the prognostic status of patients. Results We identified that smoking (HR = 2.406, 95% CI = 1.256–4.611), clinical stage III + IV (HR = 5.389, 95% CI = 2.290-12.684), the high expression group of miR-20a (HR = 4.420, 95% CI = 1.760–11.100), the high expression group of miR-197 (HR = 3.828, 95% CI = 1.778–8.245), the low expression group of miR-145 ( HR = 0.286, 95% CI = 0.116–0.709), and the low expression group of miR-30a (HR = 0.307, 95% CI = 0.133–0.706) was associated with worse prognosis. Among the five data mining models, the decision trees (DT) C5.0 model performs the best, with accuracy and Area Under Curve (AUC) of 93.75% and 0.929 (0.685, 0.997), respectively. Conclusion The results showed that the high expression level of miR-20a and miR-197, the low expression level of miR-145 and miR-30a were strongly associated with poorer prognosis in NSCLC patients, and the DT C5.0 model may serve as a novel, accurate, method for predicting prognosis of NSCLC.

show abstract

“…They have been widely used in gesture recognition applications such as hand pose estimation and gesture-based control systems. Typical decision tree algorithms are random forest regression and gradient boosting regression …”

Section: Machine Learningmentioning

confidence: 99%

“…Typical decision tree algorithms are random forest regression and gradient boosting regression. 69 3.1.7. Deep Learning (DL).…”

Section: Support Vector Machine (Svm)mentioning

confidence: 99%

Roadmap to Human–Machine Interaction through Triboelectric Nanogenerator and Machine Learning Convergence

Babu,

Mandal

2024

ACS Appl. Energy Mater.

View full text Add to dashboard Cite

Embark on a journey to unlock the pivotal roadmap for seamlessly integrating triboelectric nanogenerators (TENGs) with advanced machine learning algorithms. This review article endeavors to present a comprehensive strategy for the integration of machine learning algorithms with TENGs specifically tailored for gesture monitoring applications. The primary objective is to outline a meticulous methodology for the seamless fusion of TENG technology and machine learning techniques by elucidating the key tools for data collection, robust analysis, potential challenges, and compelling case studies that highlight the tangible applications of this fusion across diverse domains. This review not only delves into the underlying principles of TENG but also explores the fundamental tenets of machine learning, making it accessible to a wide readership, from novices to experts. The goal is to facilitate the effective implementation of this integrated framework, enabling the development of more efficient and sophisticated gesture monitoring solutions for elevating the human−machine interaction to greater heights.

show abstract

Decision trees: from efficient prediction to responsible AI

Cited by 21 publications

References 148 publications

Diagnostic and prognostic value of ECG-predicted hypertension-mediated left ventricular hypertrophy using machine learning

Diagnostic and prognostic value of ECG-predicted hypertension-mediated left ventricular hypertrophy using machine learning

Prognostic value of plasma microRNAs for non-small cell lung cancer based on data mining models

Roadmap to Human–Machine Interaction through Triboelectric Nanogenerator and Machine Learning Convergence

Contact Info

Product

Resources

About