Ensemble Flower-Pollination Algorithm for Automatic Breast Cancer Classification using Haar Features

Abstract

Breast cancer is a deadly disease which reports a higher mortality rate, every year. This type of cancer is more common in women of developed countries while compared with the incident rate in developing countries. The objective of this work is to propose a methodology to increase earlier breast cancer detection. The methodology makes use of ensemble-based classifiers namely histogram-based Gradient Boosting Machine (hbGBM), Gradient Boosting with Light GBM (lGBM), and Gradient Boosting with CatBoost (GBCB) algorithms. The paper adopts the Mammogram Image Analysis Society (MIAS) database for testing the effectiveness of the algorithms. Haar wavelet transform is used for feature extraction and Flower-Pollination Algorithm (FPA) is employed for selecting the best features. Thus, a flower-pollination based haar wavelet feature with three different ensemble-based classifiers is proposed for the automatic mammogram classification. The performance results are then compared briefly by using some benchmark metrics such as specificity, sensitivity, precision, accuracy, and precision, F1 score, and Matthews Correlation Coefficient (MCC) analysis. The result reveals that the highest classification accuracy is obtained for ensemble GBCB based FPA algorithm i.e., 91% when comparing with other existing models.