Breast Cancer Recurrence Prediction in Biopsy Using Machine Learning Framework

Abstract

There are numerous toxic and lethal substances that are available in the environment due to the rapid industrial growth and excessive application of pesticides. These substances get into the human food chain majorly through air, water, and soil. This paper presents a case study investigating the organochlorine pesticide levels in women experiencing the malignant and benign growth of breast cancer disease in order to evaluate the exposure against the chemicals and its association with the risk of breast cancer among women. After obtaining the blood and adipose tissue samples, levels of 51 chemicals including DDT, its metabolites, and isomers of HCH among 50 women each with the malignant and benign growth of breast cancer disease are measured. The levels of the chemicals in women with malignant growth of breast cancer are compared with benign cases and a prediction model is built using popular ensemble machine learning framework. The proposed framework is an optimized version of Random Forest algorithm in which feature selection is implemented and the process is incorporated with the preprocessing filters. The proposed framework for breast cancer prediction successfully achieved a prediction accuracy of 90.47%, which is found to be better than the standard classifiers like SVM, neural network, etc.