Heart-attack prediction using ML

Heart attack prediction using machine learning algorithms and pre-processing data

It's a binary classification problem (0 and 1) where for 1 the patient had a heart attack and for 0 he didn't. The dataset used was the heart attack dataset from UC Irvine. The data contains 303 samples.

My final model achieved 81% of accuracy.

Pre-processing

• Exploratory Data Analysis (EDA)
• Separate between train and test data
• Removing null values
• Removing outliers through boxplot
• Normalizing data
• PCA analysis
• Analyse the correlation matrix

Machine learning algorithms

• Dummy classifier (week baseline)
• Random Forest
• XGBoost
• Naive Bayes
• Logistic Regression
• Support Vector Machines

Metrics

Applied the following metrics to analyse the models results:

• Accuracy
• ROC Curve
• F1 Score

The best algorithm was Random Forest with 79% of accuracy.

Fine-tuning

• Random Grid search
• Cross-validation

The fine-tuned model improved the accuracy to 81%. Other papers achieved around 80% of accuracy as well.

Possible improvements

• Aggregate patients data like age by range instead of raw numbers
• Test removing variables with 0.40 correlation
• Use log(x) for the "caa" variable to normalize the data distribution
• Use other algorithms like Catboost
• Tune XGBoost hyperparameters