k-nearest neighbor algorithm

Several factors must be considered to predict the mortality of patients with hematologic malignancies. These factors and characteristics complicate doctors' and nurses' ability to predict the prognosis of such diseases. This work aimed to develop a mortality prediction model for leukemia patients utilizing artificial intelligence and a nearest-neighbor genetic algorithm. This retrospective study employed the medical records of 235 leukemia patients at Ahvaz Oncology Center from 2016 to 2019. To provide a mortality prediction model, a genetic algorithm and nearest neighbor were used. A genetic approach was employed to identify the determinants of mortality, and the nearest neighbor technique was utilized to enhance model accuracy. Ultimately, the diagnostic power of the mortality prediction model was assessed using accuracy, sensitivity, and specificity criteria. The laboratory values and variables incorporated into the genetic algorithm revealed that mechanical ventilation, hemodialysis, neutropenia, and bone marrow transplantation significantly influenced the mortality rates of patients with leukemia. The diagnostic accuracy of the genetic algorithm introduced in this study was 77.4%, with a sensitivity of 78.2% and a specificity of 82%. The results showed the artificial intelligence algorithm's effectiveness in predicting leukemia patients' mortality.