Scholarlink

<p>In recent years, the rate of cesarean sections has steadily increased, exceeding the thresholds recommended by health authorities. Despite the advances in medical knowledge, many lack access to technological tools and clinical decision support systems capable of assessing cesarean necessity based on antepartum and intrapartum factors. This research aims to mitigate these limitations by evaluating the performance of Random Forest with three prominent sampling techniques, namely synthetic minority oversampling (SMOTE), adaptive synthetic sampling (ADASYN), and random oversampling techniques on a dataset of 6157 childbirth records collected from four hospitals in Spain. The algorithms tested the suitability of the ensemble Random Forest alongside Kernel-Principal Component Analysis (Kernel-PCA) with the polynomial kernel on six distinct delivery categories: cesarean section, spontaneous vaginal delivery, forceps-assisted, vacuum-assisted, emergency cesarean, and episiotomy deliveries. Random forest (RF) combined with Kernel-PCA achieved a high F-measure of 97.64%, emphasizing the potential for integrating machine learning with clinical decision support systems to contribute to childbirth method selection.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Latent feature discovery with kernel-PCA and random forest for childbirth method classification

  • 期刊论文

  • 发表日期：

    2025年7月22日

查看全文
  • Pusarla Sindhu,
  • Parasana Sankara Rao

摘要

In recent years, the rate of cesarean sections has steadily increased, exceeding the thresholds recommended by health authorities. Despite the advances in medical knowledge, many lack access to technological tools and clinical decision support systems capable of assessing cesarean necessity based on antepartum and intrapartum factors. This research aims to mitigate these limitations by evaluating the performance of Random Forest with three prominent sampling techniques, namely synthetic minority oversampling (SMOTE), adaptive synthetic sampling (ADASYN), and random oversampling techniques on a dataset of 6157 childbirth records collected from four hospitals in Spain. The algorithms tested the suitability of the ensemble Random Forest alongside Kernel-Principal Component Analysis (Kernel-PCA) with the polynomial kernel on six distinct delivery categories: cesarean section, spontaneous vaginal delivery, forceps-assisted, vacuum-assisted, emergency cesarean, and episiotomy deliveries. Random forest (RF) combined with Kernel-PCA achieved a high F-measure of 97.64%, emphasizing the potential for integrating machine learning with clinical decision support systems to contribute to childbirth method selection.