Attribute Selection Based on a Hybrid Approach for Improving Classification of Breast Cancer Recurrence

Main Article Content

Ameer K. AL-Mashanji

Abstract

Background:


A commonly occurring disease among women worldwide is breast cancer, the second deadliest form of cancer. However, death chances are remarkably reduced when the cancer is detected and prevented at an early stage.


Materials and Methods:


The main contribution of the current study is to propose a hybrid approach to attribute selection by combining the information gain method with the correlation method and to exploit the strengths of these methods for improving classification accuracy. The dataset has been obtained from the publicly open UCI machine learning repository. The dataset is used to classify the target class into breast cancer recurrence and non-recurrence. Classification algorithms Naïve Bayes, J48 Decision Tree, and Multi-Layer Perceptron were adopted for performing the accuracy of prediction.


Results:


The proposed hybrid approach has been combined with each classification model, improving the performance of each model through the reduction of lower-ranked attributes, due to their insignificant contribution and the possibility of misguiding the classifying algorithm. After selecting a set of upper-ranked attributes carefully, it has been found that the accuracy rate, RMSE, and computational costs have improved for all three algorithms. The J48 Decision Tree achieved a significant performance, and it obtained a relatively higher accuracy (75.87 %).


Conclusions:


It can be concluded that (Inv nodes, deg-malig, node-caps, tumor size, irradiat, and breast) are strong attributes in a dataset and (Age, breast-quad, and menopause) are weak attributes. As noted, the implementation of the hybrid approach improved the accuracy of all classifiers.

Article Details

How to Cite
[1]
“Attribute Selection Based on a Hybrid Approach for Improving Classification of Breast Cancer Recurrence”, JUBPAS, vol. 31, no. 3, pp. 293–313, Sep. 2023, Accessed: Apr. 24, 2024. [Online]. Available: https://journalofbabylon.com/index.php/JUBPAS/article/view/4848
Section
Articles

How to Cite

[1]
“Attribute Selection Based on a Hybrid Approach for Improving Classification of Breast Cancer Recurrence”, JUBPAS, vol. 31, no. 3, pp. 293–313, Sep. 2023, Accessed: Apr. 24, 2024. [Online]. Available: https://journalofbabylon.com/index.php/JUBPAS/article/view/4848

Similar Articles

You may also start an advanced similarity search for this article.