Machine learning model matters its accuracy: a comparative study of ensemble learning and AutoML using heart disease prediction

What is it about?

Ensemble machine learning combines multiple models to improve performance by leveraging weak individual models. Researchers aim to enhance classification and prediction accuracy, emphasizing model quality. Precision, accuracy, and F1 scores were compared for decision tree, logistic regression, support vector machine, random forest, neural networks, Gaussian, K-nearest neighbors, and multilayer perceptron. Majority voting identified the best model for deployment. AutoML, supporting binary classification and regression, streamlines model generation without feature engineering. Comparing 18 models (8 individually trained, 10 from AutoML) on heart disease data revealed top performers: support vector, logistic regression, and neural networks (80% accuracy). AutoML recommended generalized linear (88%), gradient boosting (87%), and random forest (87%) models for optimal predictions.

Featured Image

Why is it important?

This research compares 18 machine learning models, including eight individually trained models and ten from AutoML, using an open-source heart disease dataset. Accuracy, MSE, and R² scores were evaluated. Support vector, logistic regression, and neural network models achieved 80% accuracy, outperforming Gaussian, K-nearest neighbors, and multilayer perceptron models with 76% accuracy. AutoML yielded higher accuracy, with the generalized linear model (88%), gradient boosting model (87%), distributed random forest model (87%), and extra tree model (82%) performing best.

Perspectives