Scholarlink

<p>Dementia is a progressive neurodegenerative disease that affects millions worldwide and is a challenge for early detection due to the cost, invasiveness, and limited access of available clinical methods. This study proposes an explainable Artificial Intelligence (AI) system for the prediction of risk for dementia based solely on lifestyle parameters, thus providing a non-invasive and economical option. We tested a spectrum of machine learning methods (Decision Tree, Random Forest, K-Nearest Neighbor, Naïve Bayes, Support Vector Machine, and Multilayer Perceptron) and also deep learning models (Hybrid Neural Network). Out of those, Random Forest exhibited the best accuracy of 93.1% on lifestyle data and 95.2% on an independent external validation dataset, exhibiting excellent generalizability. To combat the crucial issue of model explainability, SHapley Additive exPlanations (SHAP) was utilized for providing both the global and local interpretability and highlighting the importance of modifiable lifestyle parameters, e.g., physical activity, smoking, and comorbidities, in the determination of dementia risk. Class imbalance was overcome using methods like SMOTE, Gaussian augmentation, and scaling, thus enabling reliable detection of minority classes. Results confirm that the combination of explainable AI and efficient ML/DL methods provides accurate, interpretable, and clinically significant prediction of dementia risk. The constructed system has the potential for supporting active healthcare interventions, especially in resource-constrained environments where advanced diagnosis equipment is not available.</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

An explainable framework for dementia risk prediction based on lifestyle data using machine learning and deep learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月20日

查看全文
  • Md. Kowsar Ahmed,
  • Anichur Rahman,
  • Dipanjali Kundu,
  • Md. Alve Authi Alam,
  • Maharabe Uddin Sarker Ovi,
  • Shahariar Hossain Mahir,
  • Md Tanjum An Tashrif,
  • Abu Saleh Musa Miah,
  • Fahmid Al Farid,
  • Sarina Mansor

摘要

Dementia is a progressive neurodegenerative disease that affects millions worldwide and is a challenge for early detection due to the cost, invasiveness, and limited access of available clinical methods. This study proposes an explainable Artificial Intelligence (AI) system for the prediction of risk for dementia based solely on lifestyle parameters, thus providing a non-invasive and economical option. We tested a spectrum of machine learning methods (Decision Tree, Random Forest, K-Nearest Neighbor, Naïve Bayes, Support Vector Machine, and Multilayer Perceptron) and also deep learning models (Hybrid Neural Network). Out of those, Random Forest exhibited the best accuracy of 93.1% on lifestyle data and 95.2% on an independent external validation dataset, exhibiting excellent generalizability. To combat the crucial issue of model explainability, SHapley Additive exPlanations (SHAP) was utilized for providing both the global and local interpretability and highlighting the importance of modifiable lifestyle parameters, e.g., physical activity, smoking, and comorbidities, in the determination of dementia risk. Class imbalance was overcome using methods like SMOTE, Gaussian augmentation, and scaling, thus enabling reliable detection of minority classes. Results confirm that the combination of explainable AI and efficient ML/DL methods provides accurate, interpretable, and clinically significant prediction of dementia risk. The constructed system has the potential for supporting active healthcare interventions, especially in resource-constrained environments where advanced diagnosis equipment is not available.