Scholarlink

This study investigates the mental health of Syrian adolescent refugees aged 10 to 14 years in Amman, focusing on predicting depression symptoms based on some intervention and screening mechanism. The screening process involved household visits, where eligible adolescents and their caregivers provided consent and assent to participate. Using the Pediatric Symptom Scale (PSC-17), adolescents were screened, and those meeting the criteria were enrolled in the study. The dataset comprised 471 adolescents. Depression was assessed using the Patient Health Questionnaire (PHQ). Feature selection methods, including Gain Ratio, Gini Index, ANOVA, Chi-Squared Test, Relief-F, and FCBF, identified significant predictors of depression. Machine learning classifiers such as Decision Tree, Logistic Regression, SVM, k-NN, AdaBoost, Gradient Boosting, Naive Bayes, Random Forest, and ANN were evaluated for their predictive performance. The SVM model achieved the highest accuracy (0.832), while Naive Bayes recorded the highest AUC (0.910). Key findings highlight the importance of features like “General Functioning,” “Post-Traumatic Stress,” and “Mental Well-Being” in predicting depression. The study underscores the potential of integrating machine learning models into psychosocial programs to enhance mental health interventions for vulnerable populations.

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

Machine Learning Models for Assessing Depression in Syrian Adolescent Refugees in Jordan

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rand Habashneh,
  • Hazem Qattous,
  • Malek Alsmadi,
  • Abedalrhman Alkhateeb

摘要

This study investigates the mental health of Syrian adolescent refugees aged 10 to 14 years in Amman, focusing on predicting depression symptoms based on some intervention and screening mechanism. The screening process involved household visits, where eligible adolescents and their caregivers provided consent and assent to participate. Using the Pediatric Symptom Scale (PSC-17), adolescents were screened, and those meeting the criteria were enrolled in the study. The dataset comprised 471 adolescents. Depression was assessed using the Patient Health Questionnaire (PHQ). Feature selection methods, including Gain Ratio, Gini Index, ANOVA, Chi-Squared Test, Relief-F, and FCBF, identified significant predictors of depression. Machine learning classifiers such as Decision Tree, Logistic Regression, SVM, k-NN, AdaBoost, Gradient Boosting, Naive Bayes, Random Forest, and ANN were evaluated for their predictive performance. The SVM model achieved the highest accuracy (0.832), while Naive Bayes recorded the highest AUC (0.910). Key findings highlight the importance of features like “General Functioning,” “Post-Traumatic Stress,” and “Mental Well-Being” in predicting depression. The study underscores the potential of integrating machine learning models into psychosocial programs to enhance mental health interventions for vulnerable populations.