Hybrid Stacked Autoencoder-Graph Attention Network with Chebyshev Cloud Drift Optimization for Treatment Recommendation

Nowadays, the medical treatment recommendation model based on artificial intelligence has gained significant advancement with the increasing availability of clinical datasets and promising outcome. The traditional recommendation models have the issues with high dimensionality, redundant features and limited interpretability that negatively impact the predictive accuracy and reliability. To address the challenges, this research introduces a novel hybrid framework designed with feature optimization and deep learning-based recommendation. The proposed model is designed with preprocessing in the initial stage using one-hot encoding to handle categorical variables in the machineinterpretable format. Then, the optimal best feature selection using the Chebyshev Cloud Drift Optimization (ChCdO) is proposed for ensuring the acquisition of the most informative and relevant attributes. Finally, the proposed hybrid Stacked Auto Encoder-Graph Attention Network (SA-GAT) model is utilized for treatment recommendation using the selected attributes by the ChCdO algorithm. The experimental results demonstrate that the proposed system achieved superior accuracy and generalization performance for enhanced healthcare decision support mechanism.