EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Predicting Separation Anxiety Risk in Preschool Children Using Machine Learning Algorithms: An Approach from Variable Importance Analysis

Andrés Solano-Barliza, Wilson Arrubla-Hoyos, Rossy Jaraba-Vergara, Carmen Medina-Paternina

Investigación e Innovación en Ingenierías · 30.06.2026 · 14 (1) · 249-264

DOI: 10.17081/invinno.14.1.8902 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Investigación e Innovación en Ingenierías, online erschienen am 30.06.2026. Verfasst von Andrés Solano-Barliza und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

To identify the most influential predictors of Separation Anxiety Disorder (SAD) risk in preschool children using a reproducible machine learning framework combined with model interpretability techniques. Methodology: A quantitative predictive design was implemented utilizing a secondary dataset of 917 records of children aged 3 to 6 years. The data pipeline integrated the SMOTE technique exclusively on the training set to mitigate class imbalance without incurring data leakage, and a classifier based on the Random Forest algorithm was trained using a standard 70/30 split for training and testing. Clinical transparency and variable importance were evaluated using SHAP values.

Results

The optimal Random Forest model achieved outstanding performance on the independent test set, registering an accuracy of 0.91 and an Area Under the Receiver Operating Characteristic curve (AUC-ROC) of 0.98. However, a recall of 0.72 was identified for the risk class. The SHAP interpretability analysis revealed that avoidance of loneliness, actual distress in the absence of attachment figures, and anticipatory distress constitute the most decisive predictive variables.

Conclusions

The study validates the feasibility and accuracy of machine learning models for the early screening of psychological risks. The consistency of the variables identified by SHAP with attachment theory demonstrates the robustness of the model, although the recall level highlights the need to explore adjustments in decision thresholds for real-world clinical settings.

Zitieren

APA

Solano-Barliza, A., Arrubla-Hoyos, W., Jaraba-Vergara, R., & Medina-Paternina, C. (2026). Predicting Separation Anxiety Risk in Preschool Children Using Machine Learning Algorithms: An Approach from Variable Importance Analysis. Investigación e Innovación en Ingenierías, 14(1), 249-264. https://doi.org/10.17081/invinno.14.1.8902

Vancouver

Solano-Barliza A, Arrubla-Hoyos W, Jaraba-Vergara R, Medina-Paternina C. Predicting Separation Anxiety Risk in Preschool Children Using Machine Learning Algorithms: An Approach from Variable Importance Analysis. Investigación e Innovación en Ingenierías. 2026;14(1):249-264. doi:10.17081/invinno.14.1.8902

BibTeX
@article{SolanoBarliza2026,
  author = {Solano-Barliza, Andrés and Arrubla-Hoyos, Wilson and Jaraba-Vergara, Rossy and Medina-Paternina, Carmen},
  title = {Predicting Separation Anxiety Risk in Preschool Children Using Machine Learning Algorithms: An Approach from Variable Importance Analysis},
  journal = {Investigación e Innovación en Ingenierías},
  year = {2026},
  volume = {14},
  number = {1},
  pages = {249--264},
  doi = {10.17081/invinno.14.1.8902}
}
Kostenlos starten →