Learnable Discrete Wavelet (LDW) Pooling in CNN for Multidisciplinary Disease Prediction in Healthcare

Recommender systems are increasingly vital in healthcare, assisting in predicting crucial health information for patients and physicians alike. This study introduces an Intelligent Health Recommender System (HRS) utilizing the Learning Discrete Wavelet Pooling (LDW-Pooling) technique within a Convolutional Neural Network (CNN) framework. Unlike conventional CNN architectures, LDW-Pooling offers greater adaptability in pooling operations, resulting in enhanced accuracy and efficiency in feature extraction. The effectiveness of the proposed system is demonstrated through evaluation on a comprehensive multi-disease dataset covering heart, liver, and kidney conditions. LDW-Pooling exhibits superior accuracy in feature extraction compared to traditional pooling methods. By adaptively adjusting pooling sizes, LDW-Pooling can capture more intricate patterns and variations within medical data, leading to more accurate disease predictions Results illustrate LDW-Pooling’s superiority over traditional methods, achieving an impressive accuracy rate of 98.1%. This highlights the system’s potential to deliver precise and reliable multidisciplinary disease prediction and recommendation within healthcare environments.