ABSTRACT
Artificial intelligence and machine learning have the potential to forecast emerging dis-eases by analyzing patient medical data, offering decision-support systems for healthcare practitioners. These approaches are pivotal in minimizing diagnostic inaccuracies for phy-sicians and hold significant value for patients and healthcare organizations in scrutinizing medical records. Nevertheless, the effectiveness of algorithms fluctuates based on various attributes of datasets, fluctuations in the size of datasets, diverse parameters and their char-acteristics, and varying quantities of patient records. In this research, two distinct models have been suggested for categorizing five types of anemia diseases using the multivariate adaptive regression spline method: a 1st-degree multiclass multivariate adaptive regression spline model and a 2nd-degree multiclass multivariate adaptive regression spline model. These models were employed to classify entries into non-anemia, folate deficiency anemia, Hgb-anemia, B12 deficiency anemia, and iron deficiency anemia. When the results were analyzed, 98.30% accuracy, 92.91% precision, 92.76% recall, and 92.74% F1-score were obtained with the 2nd order multiclass multivariate adaptive regression spline model. The outcomes obtained aim to offer valuable insights to medical students and physicians en-gaged in addressing the anemia classification challenge. The application of these models holds significant promise in enhancing diagnostic procedures, minimizing error rates, and devising more efficient treatment strategies. At the same time, in addition to the studies using the same data set in the literature, it has been shown that the adaptive spline method also shows successful results.