Scholarlink

This research has two critical research objectives. The first objective is to improve the prediction model for exploring the food consumption behavior of diabetic patients among Thai people in the northern region. The second objective is to estimate the model’s performance and to extract factors that are significant to the model. The research data is a stratified random sampling of 370 people in Chiang Muan District, Phayao Province, Thailand. The data was collected between April and August 2023. Research tools include IOC-validated questionnaires, wrapper feature selection methods, and machine learning techniques using single model and ensemble machine learning techniques, including decision tree, Generalized Linear Model, K-Nearest Neighbor, Logistic Regression, Naïve Baye, and ensemble models. All models’ performance was evaluated using the cross-validation method and confusion matrix indices. The results showed that the improved and developed model had the highest efficiency and ability to predict positive questions at 65.95% and negative questions at 71.89%. Important factors from the studied model included four factors from both groups of question types. The research results can be used to drive and create strategies for controlling and predicting new cases of diabetes that may arise from food consumption behaviors.

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

Improving the Prediction Model of Food Consumption Behavior Analytics of Diabetic Patients in Northern Thailand Using Data Mining Techniques

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Pratya Nuankaew,
  • Apatcharaporn Kadkasame,
  • Kunthida Sawasit,
  • Patchara Nasa-Ngium,
  • Thapanapong Sararat,
  • Wongpanya S. Nuankaew

摘要

This research has two critical research objectives. The first objective is to improve the prediction model for exploring the food consumption behavior of diabetic patients among Thai people in the northern region. The second objective is to estimate the model’s performance and to extract factors that are significant to the model. The research data is a stratified random sampling of 370 people in Chiang Muan District, Phayao Province, Thailand. The data was collected between April and August 2023. Research tools include IOC-validated questionnaires, wrapper feature selection methods, and machine learning techniques using single model and ensemble machine learning techniques, including decision tree, Generalized Linear Model, K-Nearest Neighbor, Logistic Regression, Naïve Baye, and ensemble models. All models’ performance was evaluated using the cross-validation method and confusion matrix indices. The results showed that the improved and developed model had the highest efficiency and ability to predict positive questions at 65.95% and negative questions at 71.89%. Important factors from the studied model included four factors from both groups of question types. The research results can be used to drive and create strategies for controlling and predicting new cases of diabetes that may arise from food consumption behaviors.