Enhancing deep vein thrombosis prediction in patients with coronavirus disease 2019 using improved machine learning model

Background Deep vein thrombosis (DVT) is a significant complication in coronavirus disease 2019 patients, arising from coagulation issues in the deep venous system. Among 424 scheduled patients, 202 developed DVT (47.64%). DVT increases hospitalization risk, and complications, and impacts prognosis. Accurate prognostication and timely intervention are crucial to prevent DVT progression and improve patient outcomes. Methods This study introduces an effective DVT prediction model, named bSES-AC-RUN-FKNN, which integrates fuzzy k-nearest neighbors (FKNN) with enhanced runge-kutta optimizer (RUN). Recognizing the insufficient effectiveness of RUN in local search capability and its convergence accuracy, spherical evolutionary search (SES) and differential evolution-inspired knowledge adaptive crossover (AC) are incorporated, termed SES-AC-RUN, to enhance its optimization capability. Results Based on the benchmark set by CEC 2017 and comparative analyses with several peers, it is evident that SES-AC-RUN significantly enhances search performance compared to traditional RUN methods, even standing comparably against leading championship algorithms. The proposed bSES-AC-RUN-FKNN model was applied to predict a dataset comprising 424 cases of DVT patients, totaling 7208 records. Remarkably, the model demonstrates outstanding precision, reaching 91.02%, alongside commendable sensitivity at 91.07%. Conclusions The bSES-AC-RUN-FKNN emerges as a robust and efficient predictive tool, significantly enhancing the accuracy of DVT prediction. This model can be used to manage the risk of thrombosis in the care of COVID-19 patients. Nursing staff can combine the model's predictions with clinical judgment to formulate comprehensive treatment approaches.