Artificial Intelligence Predicts Depression in Diabetes Patients with 82% Accuracy

Strategic Insights -

Researchers at Sidi Mohamed Ben Abdellah University have conducted a comprehensive study on the use of artificial intelligence in predicting depression among individuals with diabetes. According to the study, diabetes represents a significant challenge that increases the susceptibility of individuals to develop depression, emphasizing the need for effective management of the co-occurrence of diabetes and depression. The researchers presented a comparative analysis of eight distinct machine learning algorithms to predict depression among individuals with diabetes, utilizing a dataset from Morocco and employing various techniques such as feature selection and hyperparameter tuning.

Key Takeaways:

  • The study evaluated eight machine learning algorithms, including logistic regression, k-nearest neighbors, decision tree, random forest, Adaptive Boosting, support vector machine, Extreme Gradient Boosting, and Categorical Boosting.
  • The random forest and Categorical Boosting classifiers emerged as top performers, achieving an impressive accuracy rate of 82% in predicting depression among individuals with diabetes.
  • The study utilized a dataset from Morocco, specifically curated for this purpose, and employed the Synthetic Minority Over-sampling Technique (SMOTE) to balance the dataset.
  • Feature selection was performed using ExtraTreesClassifier, while hyperparameter tuning was accomplished through a grid search approach.
  • The study conducted by Hind Bourkhime and colleagues at Sidi Mohamed Ben Abdellah University has significant implications for further research, aiming to refine prediction models in this context.
  • The researchers highlighted the considerable potential of machine learning algorithms in effectively predicting depression disorders in individuals living with diabetes.

Statistics:

  • 82% accuracy rate achieved by the random forest and Categorical Boosting classifiers in predicting depression among individuals with diabetes.
  • 8 machine learning algorithms evaluated in the study, including logistic regression, k-nearest neighbors, decision tree, random forest, Adaptive Boosting, support vector machine, Extreme Gradient Boosting, and Categorical Boosting.
  • The study obtained results showcasing promising performance of the machine learning algorithms in predicting depression among individuals with diabetes.
  • The study utilized a dataset from Morocco, specifically curated for this purpose, with 32,000 individuals with diabetes and corresponding demographic and clinical data.

Sources:

  • "Comparative analysis of machine learning algorithms for predicting depression among individuals with diabetes." Middle East Current Psychiatry, 2025, 32(1):1-9.
  • Hind Bourkhime, et al. (2025). Sidi Mohamed Ben Abdellah University Researchers Have Published New Study Findings on Machine Learning (Comparative analysis of machine learning algorithms for predicting depression among individuals with diabetes). Diabetes Week. June 30, 2025; p 230.
  • NewsRx LLC. (2025). Sidi Mohamed Ben Abdellah University Researchers Have Published New Study Findings on Machine Learning (Comparative analysis of machine learning algorithms for predicting depression among individuals with diabetes).