Anemia Prediction Using Ensemble Leaming Techniques


P.H. Menusha Prashan and K.B.N.Lakmali


Anemia, Ensemble learning, machine learning

Issue Date:

18th February 2022


Anemia can be classified as the most common blood­related disease which affects 1.62 billion people globally, which corresponds to 24.8% of the total population. The highest prevalence is in preschool-age children while the lowest prevalence is in men. Physicians use a Complete blood count (CBC) report to analyze certain parameters related to the diagnosis of anemia. This procedure belongs to the post-analytical stage of diagnosis. Due to having many variations of anemia, when predicting the anemia variant, errors can be occurred by training and non-specialized physicians. To overcome this, a system accompanied by machine Learning approach with ensemble Learning techniques is developed. To generate the classification model, boosting, stacking, bagging and voting methods are used along with classification algorithms. The system can predict the most common three types of anemias which are Thalassemia, Iron deficiency anemia and Anemia of chronic disease with the use of CBC parameters. However, the voting ensemble method performed better and achieved the highest accuracy of 0.94 % compared to other ensemble techniques with the usage of Naïve Bayes, Logistic regression, and Support Vector Machine algorithms.