Scholarlink

Suicide in Mexico is an increasing public health issue, with rates reaching 6.8 deaths per 100,000 inhabitants in 2023. This study presents a predictive system using supervised machine learning to identify individuals at suicide risk. Support Vector Machine (F1 = 99.82%), Naive Bayes (F1 = 95.41%), and Neural Networks (F1 = 99.43%) were applied to a dataset of 1,000 individuals, including variables such as past attempts, psychiatric disorders, substance abuse, anger, and social isolation. Suicidal tendencies are defined according to the DSM-5-TR as thoughts, behaviors, or intentions related to self-harm or suicide. After data preprocessing (standardization and 70/30 train-test split), models produced metrics like precision, recall, F1-score, and accuracy via a user-friendly interface. Key predictors were anger, depression, and withdrawal. The system is intended as a supportive tool to assist mental health professionals, not to replace clinical judgment. Developed in GNU Octave, it is accessible and scalable for clinical use.

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

Prediction of Suicidal Tendencies Using Machine Learning: A Multimodel Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Estefanía Luna Díaz,
  • Valeria A. Díaz Hernández,
  • Esteban T. Gardea Hernández,
  • Fernanda Solórzano Osuna,
  • Ana S. Tarango González,
  • Xóchitl Duque Alarcón,
  • Carlos Eduardo Cañedo Figueroa

摘要

Suicide in Mexico is an increasing public health issue, with rates reaching 6.8 deaths per 100,000 inhabitants in 2023. This study presents a predictive system using supervised machine learning to identify individuals at suicide risk. Support Vector Machine (F1 = 99.82%), Naive Bayes (F1 = 95.41%), and Neural Networks (F1 = 99.43%) were applied to a dataset of 1,000 individuals, including variables such as past attempts, psychiatric disorders, substance abuse, anger, and social isolation. Suicidal tendencies are defined according to the DSM-5-TR as thoughts, behaviors, or intentions related to self-harm or suicide. After data preprocessing (standardization and 70/30 train-test split), models produced metrics like precision, recall, F1-score, and accuracy via a user-friendly interface. Key predictors were anger, depression, and withdrawal. The system is intended as a supportive tool to assist mental health professionals, not to replace clinical judgment. Developed in GNU Octave, it is accessible and scalable for clinical use.