Downloads: 3 | Views: 32 | Weekly Hits: ⮙3 | Monthly Hits: ⮙3

Analysis Study Research Paper | Computer and Mathematical Sciences | India | Volume 13 Issue 3, March 2024

Integrated Feature Selection and Hyperparameter Optimization for Multi-Label Classification of Medical Conditions

Suraj Kumar ^[8] | Kukku Youseff

Abstract: The continuous evolution of data technologies in biomedical and healthcare domains has propelled the importance of precise medical data analysis for early disease recognition, improved patient care, and community services. However, the accuracy of such analyses encounters challenges in the face of incomplete medical data and regional variations in disease presentations, which can impact the precision of disease outbreak predictions. This paper addresses these challenges by investigating advanced techniques for feature selection and hyperparameter tuning to elevate the performance of a machine learning based Classifier in predicting medical conditions based on symptoms. Our methodology incorporates Recursive Feature Elimination (RFE), Mutual Information, Information Gain, and LASSO for comprehensive feature selection. The proposed approach aims make the machine learning classification more deployable in real world because the data set used has 132 symptoms for classifying 42 diseases, in real world we cannot ask for these many symptoms to a person that?s where feature engineering and hyperparameter optimization is used. Though classification of the disease has achieved exceptionally high accuracy which is a clear case of overfitting, we incorporate feature engineering for more practical machine learning model. Thus, we can address the problem of overfitting and deliver a reliable diagnosis model by this approach.

Keywords: Machine learning, Feature selection, Classification

Edition: Volume 13 Issue 3, March 2024,

Pages: 408 - 413