Anemia Detection And Severity Prediction Using Classification Algorithms With Optimized Hyperparameters, Boosting Techniques And XAI

Anemia is a blood disorder where there's a shortage of blood cells or haemoglobin. If not detected or treated promptly it can pose health risks. Detecting anemia accurately and, on time is crucial for managing and preventing complications. In this study we analyzed a sample of 364 individuals to explore how different classification algorithms perform in detecting anemia and predicting its severity. We considered Logistic regression, K-Neighbours-Classifier (KNN) Support-Vector-Machine (SVM), decision-tree and Random-forest algorithms. We fine-tuned the hyperparameters using grid and random search validation techniques to optimize their performance. Additionally, we used boosting techniques like AdaBoost, Gradient Boosting, CatBoost and XGBoost to improve accuracy. To understand the model's predictions better we employed Local Interpretable Model Explanations (LIME) and SHAP (SHapley Additive exPlanations). Among the models evaluated the random forest model's with optimized hyperparameters showed the best accuracy, in detecting anemia followed by CatBoost. In the severity prediction again random-forest turned out to give the most accurate results. This research emphasizes the significance of using classification algorithms and interpretability techniques to achieve detection of anemia and predict its severity.