Scholarlink

This work discusses supervised machine learning models’ application on personalized health mentoring using various data sets emphasizing sleep, lifestyle, and student well-being. We experimented with models like K-Nearest Neighbours (KNN), Decision Trees, Logistic Regression, Support Vector Machines (SVM), and Random Forests. All the models were stringently tested on metrics such as accuracy and F1-score. The datasets tested were “Obesity Data Set,” “Sleep Health and Lifestyle Dataset,” and “Student Lifestyle Dataset,” which each offered chief attributes like demographic, behavioural, and health metrics. Steps in preprocessing included missing data management, feature scaling, and categorial variable encoding. Although favourable results were realized—reflected in accuracy figures over 90% in multiple instances—the non-dynamic, static nature of these models limits their adaptability to real-world, dynamic health coaching scenarios. Reinforcement Learning (RL), by virtue of continuous feedback-based learning and optimization ability, is argued to be the better paradigm. This paper lays down an underlying comparison of supervised models with RL, recommending the latter in dynamic as well as user-adaptive situations.

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

Personalized Health Coaching: Classical Machine Learning vs Reinforcement Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Saswat Sharma,
  • Riddhiman Dey Sarkar,
  • Satyasaran Changdar

摘要

This work discusses supervised machine learning models’ application on personalized health mentoring using various data sets emphasizing sleep, lifestyle, and student well-being. We experimented with models like K-Nearest Neighbours (KNN), Decision Trees, Logistic Regression, Support Vector Machines (SVM), and Random Forests. All the models were stringently tested on metrics such as accuracy and F1-score. The datasets tested were “Obesity Data Set,” “Sleep Health and Lifestyle Dataset,” and “Student Lifestyle Dataset,” which each offered chief attributes like demographic, behavioural, and health metrics. Steps in preprocessing included missing data management, feature scaling, and categorial variable encoding. Although favourable results were realized—reflected in accuracy figures over 90% in multiple instances—the non-dynamic, static nature of these models limits their adaptability to real-world, dynamic health coaching scenarios. Reinforcement Learning (RL), by virtue of continuous feedback-based learning and optimization ability, is argued to be the better paradigm. This paper lays down an underlying comparison of supervised models with RL, recommending the latter in dynamic as well as user-adaptive situations.