Chulde-Fernández Bryan, Enríquez-Ortega Denisse, Guevara Cesar, Navas Paulo, Tirado-Espín Andrés, Vizcaíno-Imacaña Paulina, Villalba-Meneses Fernando, Cadena-Morejon Carolina, Almeida-Galarraga Diego, Acosta-Vargas Patricia
School of Biological Sciences and Engineering, Yachay Tech University, Hacienda San José s/n, San Miguel de Urcuquí 100119, Ecuador.
Quantitative Methods Department, CUNEF University, 28040 Madrid, Spain.
Life (Basel). 2025 Mar 19;15(3):496. doi: 10.3390/life15030496.
Several machine learning classification algorithms were evaluated using a dataset focused on heart failure. Results obtained from logistic regression, random forest, decision tree, K-nearest neighbors, and multilayer perceptron (MLP) were compared to obtain the best model. The random forest method obtained specificity = 0.93, AUC = 0.97, and Matthews correlation coefficient (MCC) = 0.83. The accuracy was high; therefore, it was considered the best model. On the other hand, K-nearest neighbors and MLP (multi-layer perceptron) showed lower accuracy rates. These results confirm the effectiveness of the random forest method in identifying heart failure cases. This study underlines that the number of features, feature selection and quality, model type, and hyperparameter fit are also critical in these studies, as well as the importance of using machine learning techniques.
使用一个专注于心力衰竭的数据集对几种机器学习分类算法进行了评估。将逻辑回归、随机森林、决策树、K近邻和多层感知器（MLP）得到的结果进行比较，以获得最佳模型。随机森林方法的特异性为0.93，曲线下面积（AUC）为0.97，马修斯相关系数（MCC）为0.83。准确率很高；因此，它被认为是最佳模型。另一方面，K近邻和MLP（多层感知器）的准确率较低。这些结果证实了随机森林方法在识别心力衰竭病例方面的有效性。这项研究强调，在这些研究中，特征数量、特征选择和质量、模型类型以及超参数拟合也至关重要，同时也凸显了使用机器学习技术的重要性。
