Prediction of breast cancer through fast optimization techniques applied to machine learning

This paper studies new accelerated optimization algorithms and applies the algorithms to the prediction of breast cancer through a machine-learning approach. We first introduce new fast CQ algorithms and obtain weak convergence results to do this. In one of our proposed algorithms (inertial-type CQ Algorithm), the inertial choice could be negative and even greater than 1 with no on-line rule imposed to obtain convergence results. This is a major improvement over other inertial-type algorithms in the literature where inertial choices are restrictive to [0,1) and on-line rule is imposed. Then we validate the applicability of the proposed CQ algorithms to real-life applications by predicting breast cancer by updating the optimal weight in machine learning. We use the mammographic mass dataset from the UC Irvine machine learning repository available on the UCI website as a training set to show the superiority of our algorithms over existing ones in the literature.