Predictive Modeling Approach for Breast Cancer Detection using Machine Learning

Abstract

The major problem for the health systems of any developed or developing nation is not newlineonly managing individuals with single chronic disease, but also people with co-occurring newlinediseases such as high blood pressure, heart disease, and diabetes. If the spread of breast newlinecancer among women is concerned, then it is the majorly detected cancer amongst the newlinewomen especially of the younger age and it holds the second position worldwide. As a newlineresult, an immense rise in incidence and mortality can be seen. It is rising at a rapid rate newlineand holds 25% amongst all types of cancer and a deceased rate of 15% worldwide. If it is newlinediagnosed at an early stage, then the chances of getting cured are up to 100%, but if it is newlinediagnosed at the last stage then it can be more harmful and the chances of survival get newlinereduced to 15%. Machine learning techniques like classification, clustering, feature newlineselection, etc. can be helpful in this regard for early detection of breast cancer, this will help newlinein reducing the rate of mortality and strengthen the health systems of every nation across newlinethe globe. Information and Communication Technologies can provide a proper channel for newlinecommunication through which peoples can access, store, process, and transmit the data by newlineusing a fully digital form of environment which includes various electronic systems and newlineaudiovisual tools. These technological advancements have a strong potential in order to newlineprovide improved healthcare facilities, information exchange regarding the health and newlinepersonalized medicinal advice. newline newline The objective of this research is to develop a predictive model using various classification newlineand clustering algorithms which can be helpful to predict the occurrence of breast cancer newlineat an early stage. A detailed analysis has been provided regarding various parameters of newlineeffectiveness and efficiency like accuracy, True Positive Rate, False Positive Rate, Root newlineMean Square Error, precision, recall, etc. in the different chapters of this thesis which has newlinebeen used for generating the results. The defined objectives of this study have been newlineelaborated from chapter-2 to chapter-6, in which we have applied predictive modeling newlinetechniques by using classification, clustering and feature selection, as a result the accuracy newlinehas been enhanced in comparison with existing literature. Furthermore, the concept of RNN newlineand MLP have also been used for enhancement of accuracy in classification.