AN AI-ENABLED HYBRID DEEP LEARNING FRAMEWORK FOR CARDIOVASCULAR DISEASE DIAGNOSIS AND HEALTHCARE RECOMMENDATIONS

Globally, diabetes is a well-known problem that affects both developed and developing nations. The term “healthcare monitoring” describes the ongoing evaluation of a patient’s health and the recording of numerous critical symptoms. By facilitating more precise diagnosis and individualized treatment regimens, artificial intelligence (AI) is transforming the healthcare industry. The integration of AI technologies into medical frameworks enhances the ability to monitor patients continuously and assess complex health data. This paper suggests a new hybrid deep learning (DL)-based disease prediction system for healthcare monitoring and recommendations, with a focus on cardiovascular disease (CVD). Improved [Formula: see text]-score normalization is used to preprocess the input data in order to improve its reliability and quality. The process of extracting entropy features is proposed using the Biased Weighted Sample Variance-Fuzzy Entropy (BWSV-FE)-driven feature extraction method. Alongside this feature, mutual information and raw features are collected to reduce data dimensionality. Subsequently, a hybrid illness prediction model is introduced, incorporating the Long Short-Term Memory (LSTM) and Improved Bidirectional Gated Recurrent Unit (Improved Bi-GRU) models that received the extracted feature as input. For more accurate illness prediction, an Improved Bi-GRU is employed, which integrates the penalty function into the Categorical Cross-Entropy (CCE) and updates it using a modified Mean Squared Error (MSE) loss function. Additionally, the penalty is calculated using the Gini Index. Consequently, the hybrid illness prediction model effectively forecasts the severity of the disease, aiding physicians in managing patient care. Additionally, the efficacy of the suggested strategy is compared to other models that are currently in use. The Improved Bi-GRU[Formula: see text] LSTM retained an optimum specificity and accuracy of 93.7% and 94.15% at 80% of the training data, whereas other methods received lower ratings. This AI-driven framework not only enhances the precision of healthcare monitoring but also provides scalable solutions for effective individual and systemic healthcare management, facilitating better disease management and patient care.