A novel hybrid sparse-neural joint adjustable random forest (Sn-Arf) classification method for medical decision support system

A patient’s medical history, current treatment choices, and medical expertise must all be carefully considered during the complicated process of making medical decisions. To assist healthcare workers in making timely and well-informed judgments, a robust Medical Decision Support System (MDSS) is a precious resource in this regard. Medical professionals often use complicated medical data, while conventional techniques require an extensive amount of time, are highly susceptible to flaws, and can fail to provide accurate diagnoses. The identification and provision of reliable information for patient therapy need a more practical approach. The classification procedure is carried out using a Sparse-Neural Joint Adjustable Random Forest (SN-ARF). Feature selection serves as the input for the classification process, enhancing its performance. Initially, the input data are preprocessed by using data cleaning and oversampling methods. Feature selection is performed by applying Swarm Optimised Cluster Genetic Algorithm (SOC-GA). The selected features are then fed as input for classification using SN-ARF model. Finally, the classification results demonstrate excellence in accuracy (95%), precision (91%), recall (95%), and F1 score (93%), thereby indicating the high overall performance achieved by our proposed method. We compare these metrics with the conventional approaches to evaluate the detection of hazards. For healthcare professionals, SN-ARF is a potent tool that makes decision-making in a variety of medical circumstances more reliable and accurate. As a result of the system’s exceptional accuracy in medical diagnosis, treatment planning, and prognosis prediction, patient outcomes and healthcare efficiency are enhanced.