Heart Disease Prediction Using Bi-GRU with GWO for Tuning Parameter

Machine learning has improved its precision and accuracy in
classifying clinical cardiac disease datasets in recent years.
However, studies show that the quality of the heart disease feature
used for training has a substantial effect on the prediction model's
effectiveness. This work presents a reliable model for predicting
cardiac disease using data from the UCI Machine Learning
Repository. Class imbalance is addressed by increasing minority
class representation with the Synthetic Minority Upper Model
Technique (SMOTE). To pick the most relevant features and
boost the model's efficiency, a hybrid feature engineering strategy
combines forward feature selection with recursive feature
elimination (RFE). Gray Wolf Optimization (GWO) is used to
change hyperparameters for optimal model performance. A
bidirectional cadet recurrent unit (bi-GRU) is used for
classification in order to identify biases and hierarchical patterns
in the data. Accuracy, specificity, and sensitivity are used to
evaluate the model, giving a thorough evaluation of its prediction
power. A very accurate and dependable technique for predicting
heart illness is produced by combining data imbalance, improved
feature selection, optimization, and deep learning. This shows
promise for useful clinical decision support in risk assessment
and early diagnosis. The proposed approach produced the result
as sensitivity, specificity and accuracy as 91%, 90% and 98%.