Comparative Analysis of Machine Learning Models for Chronic Disease Prediction - A Multimodel Study On Diabetes, Hypertension, and Stroke

Aug 6, 2026

Upload by Eliana

This thesis applies supervised machine learning to predict three common chronic diseases—diabetes, hypertension, and stroke—using real-world clinical and health data. A complete modeling pipeline is built, including data preprocessing, feature engineering, transformation, collinearity reduction, and class balancing. Logistic Regression, KNN, Random Forest, and XGBoost are trained with cross-validation and holdout test sets. XGBoost achieves the best overall performance, notably recall and AUC, while Logistic Regression offers a strong interpretable baseline.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages39

Reading Time1 h 38 min

AccessFree to download

Introduction

Background and Motivations

Research Objectives

Data Preprocessing and Exploratory Data Analysis

Data Source

Variables Explanation

Feature Selection

Skewness Check and Transformation

Variance Inflation Factor

Modeling and Evaluation

Logistic Regression

K-Nearest Neighbors

Random Forest

XGBoost

Model Comparison

Feature Importance

Conclusion

Summary

Future Work

References

UCLA UCLA Electronic Theses and Dissertations Title Comparative Analysis of Machine Learning Models for Chronic Disease Prediction: AMultimodel Study On Diabetes, Hypertension, and Stroke Permalink [https://escholarship.org/uc/item/0jq522zw](https://escholarship.org/uc/item/0jq522zw) Author Wang, Jiaheng Publication Date 2025 Peer reviewed|Thesis/dissertation [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California UNIVERSITY OF CALIFORNIA Los Angeles Comparative Analysis of Machine Learning Models for Chronic Disease Prediction: A Multimodel Study On Diabetes, Hypertension, and Stroke A thesis submitted in partial satisfaction of the requirements for the degree Master of Applied Statistics and Data Science by Jiaheng Wang 2025 © Copyright by Jiaheng Wang 2025 ABSTRACT OF THE THESIS Comparative Analysis of Machine Learning Models for Chronic Disease Prediction: A Multimodel Study On Diabetes, Hypertension, and Stroke by Jiaheng Wang Master of Applied Statistics and Data Science University of California, Los Angeles, 2025 Professor Guang Cheng, Chair This paper exemplifies how supervised machine learning can be used to predict three common chronic diseases (diabetes, hypertension, and stroke) utilizing clinical and health data from the real world. A comprehensive modeling pipeline was developed, encompassing data preprocessing, feature engineering, transformation, collinearity reduction, and class balancing. Logistic Regression, KNN, Random Forest, and XGBoost are considered as modeling techniques and used both cross-validation and holdout test sets. XGBoost demonstrated superior performance compared to other models, especially recall and AUC, while Logistic Regression served as a strong, interpretable baseline. Hypertension models achieved near-perfect results—likely due to clear class separability confirmed by PCA—and tree-based models improved stroke prediction by capturing complex nonlinear relationships. Within the predictor variables, many variables appeared within the models (regardless of the disease) and demonstrated commonality, supporting the development of possible integrated screening and pragmatic public health intervention with evidence to show how the data drove decisions. The thesis of Jiaheng Wang is approved. Hongquan Xu Yingnian Wu Guang Cheng, Committee Chair University of California, Los Angeles 2025 TABLE OF CONTENTS 1 Introduction ...................................... 1 1.1 Background and Motivations ........................... 1 1.2 Research Objectives ................................ 2 2 Data Preprocessing and Exploratory Data Analysis ............. 4 2.1 Data Source .................................... 4 2.2 Variables Explanation .............................. 4 2.3 Feature Selection ................................. 5 2.3.1 Correlation Heatmap ........................... 6 2.3.2 Feature Importance ............................ 7 2.3.3 Full Feature Ranking ........................... 8 2.3.4 Feature Dropped ............................. 8 2.4 Skewness Check and Transformation ...................... 9 2.4.1 Log1p Transformation .......................... 10 2.4.2 Yeo-Johnson Transformation ....................... 10 2.4.3 Transformation Comparison ....................... 11 2.5 Variance Inflation Factor ............................. 12 3 Modeling and Evaluation .............................. 14 3.1 Logistic Regression ................................ 14 3.2 K-Nearest Neighbors ............................... 15 3.3 Random Forest .................................. 15 3.4 XGBoost ...................................... 16 3.5 Model Comparison ................................ 17 3.5.1 Model Performance Metrics ....................... 17 3.5.2 Problem facing and Model Selection ................... 18 3.5.3 Model Selection .......................

FAQs

What diseases are modeled in this thesis?

The thesis predicts diabetes, hypertension, and stroke using clinical and health data collected from real-world settings.

Which machine learning models are compared?

Logistic Regression, KNN, Random Forest, and XGBoost are evaluated using both cross-validation and holdout test sets.

Why does XGBoost perform best?

XGBoost shows superior recall and AUC, while the Hypertension results are near-perfect, attributed to clear class separability confirmed by PCA and tree-based models capturing complex nonlinear patterns for stroke prediction.

Aug 6, 2026

Upload by Eliana