BREAST CANCER PREDICTION WITH HYBRID ML MODELS

Abstract

In recent times various types of cancer propagation in humans are alarmingly increasing and especially women are prone to and threatened by breast cancer with high morbidity and mortality. The absence of robust prognosis models results in difficulty for physicians to prepare a treatment plan that may extend patient survival chances and time. Hence, the need of the time is to develop the technique which offers minimum error with increased accuracy. Different legacy algorithms like SVM, Regression, are compared with the proposed hybrid prediction model outcome. All experiments are executed within a parallel environment and conducted in anaconda python platform with relevant libraries. This is helpful in domains like. prediction of cancer before diagnosis, prediction of diagnosis and outcome during treatment. The proposed work combining detailed pre-prepressing stages over a deep neural network model with tuned hyper parameters, validated to yield needed accuracy. This can be used to derive and compare the outcome of different techniques and suitable one having max accuracy and stability, can be used depending upon requirement. Different data sets are tried and analysed for prediction with different parameters and results are compared. Keywords — Breast Cancer detection, machine learning, feature selection, classification, hybrid deep learning, image classification, KNN , Random Forest, ROC.