Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Chronic Disease Diagnoses Using Behavioral Data

Di Wang, Yidan Hu, Eng Sing Lee, Hui Hwang Teong, Ray Tian Rui Lai, Wai Han Hoi, Chunyan Miao

TL;DR

This work tackles the challenge of early 3H detection (diabetes, hyperlipidemia, hypertension) using daily behavioral data rather than clinical measurements. It introduces the S3D dataset collected from 629 participants over 3 months and develops a preprocessing plus feature-engineering pipeline that captures temporal changes through dynamic features like $SBP_F$, $SBP_L$, and $BG_C$, with imputation via Mean Imputation and $k$-Nearest Neighbor Imputation. Employing ML models such as XGBoost and Random Forest, it achieves accuracies of 80.2% for diabetes, 71.3% for hyperlipidemia, and 81.2% for hypertension, with SHAP analyses confirming clinically plausible influential features. The results demonstrate the feasibility and potential of home-based, behavior-driven screening to complement traditional testing, and provide explainable insights through SHAP for trust and adoption in clinical and public-health contexts.

Abstract

Early detection of chronic diseases is beneficial to healthcare by providing a golden opportunity for timely interventions. Although numerous prior studies have successfully used machine learning (ML) models for disease diagnoses, they highly rely on medical data, which are scarce for most patients in the early stage of the chronic diseases. In this paper, we aim to diagnose hyperglycemia (diabetes), hyperlipidemia, and hypertension (collectively known as 3H) using own collected behavioral data, thus, enable the early detection of 3H without using medical data collected in clinical settings. Specifically, we collected daily behavioral data from 629 participants over a 3-month study period, and trained various ML models after data preprocessing. Experimental results show that only using the participants' uploaded behavioral data, we can achieve accurate 3H diagnoses: 80.2\%, 71.3\%, and 81.2\% for diabetes, hyperlipidemia, and hypertension, respectively. Furthermore, we conduct Shapley analysis on the trained models to identify the most influential features for each type of diseases. The identified influential features are consistent with those reported in the literature.

Chronic Disease Diagnoses Using Behavioral Data

TL;DR

This work tackles the challenge of early 3H detection (diabetes, hyperlipidemia, hypertension) using daily behavioral data rather than clinical measurements. It introduces the S3D dataset collected from 629 participants over 3 months and develops a preprocessing plus feature-engineering pipeline that captures temporal changes through dynamic features like , , and , with imputation via Mean Imputation and -Nearest Neighbor Imputation. Employing ML models such as XGBoost and Random Forest, it achieves accuracies of 80.2% for diabetes, 71.3% for hyperlipidemia, and 81.2% for hypertension, with SHAP analyses confirming clinically plausible influential features. The results demonstrate the feasibility and potential of home-based, behavior-driven screening to complement traditional testing, and provide explainable insights through SHAP for trust and adoption in clinical and public-health contexts.

Abstract

Early detection of chronic diseases is beneficial to healthcare by providing a golden opportunity for timely interventions. Although numerous prior studies have successfully used machine learning (ML) models for disease diagnoses, they highly rely on medical data, which are scarce for most patients in the early stage of the chronic diseases. In this paper, we aim to diagnose hyperglycemia (diabetes), hyperlipidemia, and hypertension (collectively known as 3H) using own collected behavioral data, thus, enable the early detection of 3H without using medical data collected in clinical settings. Specifically, we collected daily behavioral data from 629 participants over a 3-month study period, and trained various ML models after data preprocessing. Experimental results show that only using the participants' uploaded behavioral data, we can achieve accurate 3H diagnoses: 80.2\%, 71.3\%, and 81.2\% for diabetes, hyperlipidemia, and hypertension, respectively. Furthermore, we conduct Shapley analysis on the trained models to identify the most influential features for each type of diseases. The identified influential features are consistent with those reported in the literature.
Paper Structure (23 sections, 2 equations, 2 figures, 2 tables)

This paper contains 23 sections, 2 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Machine Learning Models Used for 3H Diagnoses
  4. 3H Disease Diagnosis Datasets
  5. Daily Behavioral Dataset for 3H Diagnosis
  6. Using S3D Dataset for 3H Diagnosis
  7. Data Preprocessing
  8. Data Cleaning
  9. Feature Engineering
  10. Missing Value Imputation
  11. One-hot Encoder
  12. Machine Learning Algorithms Used for 3H Diagnosis
  13. Random Forest (RF)
  14. XGBoost (XGB)
  15. k-Nearest Neighbors (KNN)
  16. ...and 8 more sections

Figures (2)

  • Figure 1: Comparison of accuracy for hypertension diagnosis. ER denotes the expert rule (see Section \ref{['subsec:er']}), and MV denotes missing values.
  • Figure 2: SHAP summary plot on hyperlipidemia and hypertension diagnoses using XGB.