A Novel Weighting Attribute Method for Binary Classification

In conventional binary classification algorithms, all features are treated as having equal weight and classification models are built without taking into consideration the fact that different attributes can have different levels of influence on a class. Attribute weighting adjustments are used in machine learning models to improve performance. In this paper, we propose a novel attribute weighting method based on mutual information and apply this method to four classical machine learning models for classification. We study the performance of our weighting method by conducting experiments on the Wisconsin Breast Cancer database and Blood transfusion service center dataset. In three of the four machine learning models, our weighted attribute model outperformed the corresponding conventional machine learning models in binary classification.