Heart Disease Predication Extreme Gradient Boosting with Hyperparameter Tuning (XGBoost + Optuna/Hyperopt)

Abstract

The heart diseases have ranked among the top fatal causes in the world and therefore, the significance of the early and accurate mechanisms of identifying the diseases. The work was dedicated to the applied application of the Extreme Gradient Boosting (XGBoost) as an extremely strong method of machine learning and applying it to forecast the presence of the heart disease according to the presence of many dif- ferent clinical and a demographical variable. Comparison of Performance of Different ModelThe data being adopted into the study is imported through Kaggle and this data contains the following properties; age, Gender, type of chest pain, rest blood press, serum cholestrol, fasting blood sugar, result of an electrocardiogram, maximal heart rate, exer- induced angina, ST -segment depression and the slope of ST part and the number of vessel as well as the type of thalassemia. The dependent variable will show those who lack and have heart diseases.
Optuna and Hyperopt will be used to tune the hyperparame- ters so that the models become predictive using advanced tech- niques. Such libraries use Bayesian optimization algorithms that assist in exploring the hyperparameter space effectively and landing model configurations giving the best results. In addition to its impact, upon the model accuracy, the utilization of this fine-tuning strategy also leads to the augmentation of the capacity of this model to work well on the raw data.
According to common classification scores (precision, re- call, and F1-score), the XGBoost model is checked. Also, an analysis of feature importance is carried out to know which variables are the most I find in mapping heart disease. The findings show that the model based on XGBoost is highly functional in predicting individuals with risk of heart decease.
This predictive model can be applied practically in the clinical practice that will result in timely medical intervention and improved patient outcomes due early diagnosis. This work combines machine learning into the domain of healthcare analytics and further advances the domain of data-driven diagnostic of health outcomes.