Prediction and analysis of heart disease data based on machine learning
School of Global Public Health, New York University
Haoping Li
2023-04-18
CONTENTS
Introduction and Significance
What do we know about this problem?
Heart disease is a serious threat to human life
What is the significance?
To prove machine learning could be used to predict
Gaps
Outdated， limited to a certain region
Research Question
Research question
Hypothesis
What factors can help people better identify whether heart disease will occur by analyzing people‘s research data, and using machine learning to identify the diagnostic risk of heart disease? And find out which model would perform better among different machine-learning methods.
We assume that logical regression would be the best algorithm to predict the risk of a heart attack in this study.
Methods
Data Source and Study Design
Measures
Outcome :
Target – Diagnosis of heart disease(1 = true; 0 = false)
Exposure:
cp: chest pain type trestbps: resting blood pressure,
Statistical Analysis
Data Processing - Missing value
Training and Modeling - Training and Testing data
Model Evaluation – AUC & ROC, MSE, Confusion Matrix
Heart Disease Data Set from the UC Irvine Machine Learning Repository
Logical Regression, Decision Tree, Random Forest, KNN
Results
According to the statistics of the age variable and cholesterol variable, it can be seen from the results that whether heart disease will occur is highly related to these two factors.
The AUC value of the Logistic model is the best, indicating that the performance of the model is relatively good.
The accuracy of the random forest model is the highest, indicating that the model has the highest prediction result on the test set.
The visualization of correlation analysis refers to the analysis of two or more variable elements with correlation.
A confusion matrix is a table that summarizes the performance of a binary classification model by comparing the predicted labels with the true labels. It can help you to calculate evaluation metrics, identify types of errors, and visualize model performance.
Conclusions
Interpretation of the main finding
Limitations
The best AUC value among them is for the logistic model, which shows that the model performs rather well. The random forest model has the maximum accuracy, which means that it has the best test set prediction outcome.
Data size and time
Diversity in heart disease
The setting of the parameter
Model | Accuracy | AUC |
Logistic model | 0.1758 | 0.821 |
KNN | 0.5934 | 0.592 |
Decision tree | 0.6044 | 0.600 |
Random forest | 0.7363 | 0.734 |
Implications/Recommendations
References
Ali Md Mamun,Paul Bikash Kumar,Ahmed Kawsar,Bui Francis M.,Quinn Julian M.W.,Moni Mohammad Ali. Heart disease prediction using supervised machine learning algorithms: Performance analysis and comparison[J]. Computers in Biology and Medicine,2021,136.
Jae Won Choi,Young Keun Choi. The Prediction of Heart Disease using Machine Learning Technique[J]. International Journal of Innovative Technology and Exploring Engineering (IJITEE),2020,9(5).
Prediction of Heart Disease using Machine Learning[J]. International Journal of Recent Technology and Engineering,2019,8(210).
THE END
THANKS