Deep Ensemble Learning Approach for Robust Disease Classification Using Heterogeneous Healthcare Data

Abstract

The increased heterogeneous healthcare data, such as structured electronic health records (EHR), medical imaging, and unstructured clinical text, provide a big opportunity to enhance the classification of diseases. Nonetheless, the inconsistency of data forms, noise, and missing data restrict the usefulness of traditional machine learning methods. This paper presents the proposal of an effective deep ensemble learning framework to classify disease based on multi-modal healthcare data. The framework combines the Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and the Transformer-based architectures using the stacking ensemble approach to learn spatial, temporal and contextual characteristics. The suggested model is tested using benchmark healthcare datasets with elaborate experimental configuration, various performance measures, and statistical testing. The findings prove that the ensemble methodology works better than single models, with an accuracy of 0.91, F1-score of 0.90 and area under the curve (AUC) of 0.93. The soundness and external validity of the model are supported by statistical analysis, which involves confidence interval and significance test (p < 0.05). As indicated in the findings, deep ensemble learning is an efficient method of tackling the issues of heterogeneous medical datasets and improving classification accuracy. This is a scalable and flexible framework of clinical decision support systems and intelligent healthcare in the future.