Gaussian Light Gradient Boost Ensemble Decision Tree Classifier for Breast Cancer Detection

Abstract

Detection of cancer in the breasts shows an important role in minimizing the mortality rates and increasing the cure rate, relieve as well as guarantee the patient’s life quality. Several works have been done in the breast cancer detection but it failed to perform accurate detection with minimum time. In order to improve breast cancer detection, an ensemble technique called Gaussian light gradient boost decision tree classification (GLGBDTC) is introduced. Initially, images are collected from the database. The Light Gradient Boost technique further constructs a number of base classifiers namely c4.5 decision trees using Kullback–Leibler divergence value, by which the data are classified and the results are to be sum up for making strong classification outcomes. For all the base classifiers, the similar weights are assigned. Then the Gaussian training loss is computed for each base classifier results. Followed by, the weight is updated according to the loss value. The steepest descent function is used to discover best classifier with minimum training loss. By this way, the proposed technique performs accurate breast cancer detection. The simulation results show minimize false positive rate (FPR).