Written below is the overall flow of our presentation and an overview of our work. More details can be found in the comments in our notebook and we talk through our whole process in our video presentation.
Our project aims to see if machine learning algorithms can accurately predict for cardiovascular disease and tell us what are the factors that contribute most to the presence of cardiovascular disease.
The dataset that we used was this
We utilized methods we learnt during the Data Science section of the course to observe any distinctive patterns across the 2 different cardio classes
From observation, we were able to come up with some hypothesis that certain variables will play a bigger role in determining for the presence of cardiovascular diseases compared to other variables.
This helped our model to see a more diverse combination of data and helped eliminate any potential overfitting and information leakage during hyperparameter tuning.
We observed the feature importances of our trained and tuned models to check our initial hypothesis formed from our EDA to decide which features play a role in determining for cardiovascular disease.
Additionally, we also look at the confusion matrix and realize that our models have further room for improvement.
- Logistic Regression
- Decision Tree Classifier
- Random Forest Classifier
- XGBClassifier
- KNeighborsClassifier