Vitamin deficiency prediction using metaheuristic algorithms​

Blood has held significant importance throughout human existence. Even in ancient times, people recognized the crucial role of blood in determining diseases and health conditions. With advancements in the field of medicine and technological progress, blood tests have become a vital tool for understanding individuals' health. Therefore, blood analysis has become an integral part of medical diagnosis and treatment processes. In this study, a project was conducted on the prediction of B12 vitamin levels. The dataset includes blood values of a total of 907 individuals, ranging from 2 to 92 years, with 509 females and 398 males. Parameters of models used to enhance the performance of regression models and reduce the risk of overfitting in SVR, LASSO, and Ridge regressions were optimized. For SVR, the best 'C' and gamma values were found to be 0.072 and 0.107, respectively, resulting in improved outcomes with MAE, MSE, RMSE, and 𝑅2 values of 0.145, 0.024, 0.152, and 0.890 after optimization. The optimal 'alpha' value for Lasso regression was 0.011, leading to enhanced results post-optimization with MAE, MSE, RMSE, and 𝑅2 values of 0.030, 0.003, 0.062, and 0.983. In Ridge regression, the best 'alpha' value was determined as 0.013, and the results improved after optimization with MAE, MSE, RMSE, and 𝑅2 values of 0.063, 0.079, 0.289, and 0.623. This study emphasizes the importance of blood analysis in the medical field and the predictive power of regression analysis.