Analyzing Factors Affecting Cholesterol Levels: A Data-Driven Study Using Statistical Models and Machine Learning

Abstract

High cholesterol levels are associated with various health complications, particularly cardiovascular diseases. Predicting an individual's cholesterol levels can be crucial in healthcare analytics to prevent and manage these conditions, leading to long-term health benefits and potential economic savings in healthcare systems. This paper focuses on analyzing the factors influencing blood cholesterol levels and developing predictive models for cholesterol levels using statistical and machine learning techniques. The study conducted an extensive analysis of determinants governing cholesterol levels using a feature importance ratio and leveraged the Framingham Heart Study (FHS) dataset with machine learning techniques to predict cholesterol levels. The study found that age, BMI, and glucose levels consistently influenced cholesterol levels, whether classified into three levels or two levels. Machine learning models exhibited varying performance, with models like Random Forest and Gradient Boosting excelling in precision, recall, and F1-score in specific cholesterol categories. The results emphasize the importance of addressing age, BMI, and glucose levels in healthcare strategies for cholesterol management. They also highlight the need for continuous model refinement and fine-tuning to improve predictive accuracy in different cholesterol classification scenarios.