Robust Heart Disease Prognosis: Integrating Extended Isolation Forest Outlier Detection with Advanced Prediction Models

Heart disease can be prevented with an accurate prognosis, but it can also be catastrophic if the forecast is erroneous. This paper presents an innovative approach for the prediction of heart disease by using machine learning and deep learning techniques with Extended Isolation Forest outlier detection. The proposed system leverages a dataset comprising various clinical features and diagnostic parameters associated with heart disease. Initially, the extended isolation forest algorithm is employed to identify outliers and mitigate their influence on subsequent analyses. This preprocessing step enhances the overall robustness of the prediction system. Next, machine learning and deep learning models is utilized for the prediction of heart disease. Multiple classification algorithms, including logistic regression, support vector machine, random forest and light gradient boosting are trained on the preprocessed dataset to identify patterns and relationships between the features and disease outcomes. Furthermore, a deep learning model, such as gated recurrent unit is employed to extract intricate patterns from the input data and capture temporal dependencies. The accuracy and confusion matrix is used to validate several promising outcomes. The integration of outlier detection techniques further enhances the system’s performance by minimizing the impact of erroneous data, and also the data has been standardized to achieve optimal results. The accuracy of 93.4% was achieved using a deep learning method.