HEART DISEASE PREDICTION USING RNN WITH ADVANCED PREPROCESSING AND FEATURE EXTRACTION TECHNIQUES

Abstract

One of the biggest causes of death worldwide is still heart disease, which emphasizes the need for reliable predictive models to support early detection and treatment. Developing precise and effective prediction systems is essential since conventional approaches for diagnosing cardiac disease sometimes rely on difficult and time-consuming processes. By utilizing huge datasets and complex algorithms, ML(Machine Learning) techniques have become increasingly effective tools for improving diagnostic procedures in recent years. In this work, we used a Recurrent Neural Network (RNN) in conjunction with extensive preprocessing and feature extraction techniques to construct an advanced system for heart disease prediction. Min-max scaling, IQR (Interquartile Range), and z-score normalization were used in the preprocessing stage to standardize the data and guarantee consistent input for the RNN model.  Chi-square tests to extract features, substantially reduced the number of dimensions and improved the relevance of the features. The heart disease dataset from the UCI ML Repository was used to assess the suggested system. The results showed that model performance was greatly enhanced across accuracy, precision, and recall measures when IQR and Chi-square tests were combined with RNN.