Optimized Hybrid Feature Selection Techniques for Detecting Iron Deficiency Anemia

Abstract

The Iron Deficiency Anemia (IDA) is one of the most common types of nutritional disorders in the world and it requires precise and timely diagnosis to avoid the consequences of its development in the human body. This work aims is to improve and boost the classification performance of diagnosing IDA by utilizing different Feature Selection Techniques (FST) on the basis of filter, wrapper, embedded and hybrid approaches. A dataset containing the biological markers was compiled for analysis and several algorithms like Analysis of Variance (ANOVA) F-statistic, Recursive Feature Elimination (RFE), Least Absolute Shrinkage and Selection Operator (LASSO), Mean Squared Error (MSE), Random Forest and Support Vector Machine (SVM) from the above FST were used to determine the most discriminative features. Also, some hybrid algorithms from statistical and model-based selection, including ANOVA with Logistic Regression (Anolog) and Random Forest with Chi-square (ChiForest) were developed and evaluated. Based on their performance, the most valuable features were selected and thus the performance evaluation is enhanced. This comprehensive study highlights the effectiveness of hybrid feature selection methods to enhance the diagnostic accuracy, the model efficiency and clarity of interpretation. It is suggested by the findings that advanced machine learning and feature selection techniques should be integrated to come up with robust diagnostic tools that could be used to identify IDA.

Keywords: Iron Deficiency Anemia(IDA), Feature Selection Techniques(FST), Filter, wrapper and Embedded methods, Hybrid feature selection techniques.