Design and Development of Efficient Classification Models to Predict Breast Cancer

Abstract

Breast Cancer is the most serious, prevalent, and menacing Cancer pertaining to women. newlineWith the privation of robust prognosis models, doctors face difficulty to prepare a proper treatment newlineplan that persists patient survival duration. Hence, with the advancement of modern techniques newlineand prediction models, this problem can be addressed to maximum accuracy. Machine learning is newlineone such tool, which acts as a subset of artificial intelligence, focusing on using statistical data to newlinebuild an intelligent system by training the available database. newlineAutomated Machine Learning Models can be used in the early prediction of Breast Cancer, newlinewhich can enhance the survival rate of women suffering from this dreadful disease. Proposed newlineresearch work has been carried out using the UCI repository, Kaggle and BreakHis static newlinerepositories dataset. The researcher in this work deals with the development of the classification newlinemodel after applying the feature selection to get better accuracy, developing a unique Neural newlineNetwork Model to predict the proximity of Breast Cancer. The model is trained by an improvised newlinefeature selection technique that enhances the accuracy of the classification model. newlineThis study initiates with the first contribution by the estimation of results using feature newlineselection techniques such as Correlation-based and Information-gain based Ranker algorithm, newlinethese algorithms select the best subset from the overall attributes that can be passed to the five newlineclassifiers (Sequential Minimum Optimization, Multilayer Perceptron, K-Star, Decision Table, newlineand Random Forest) to predict the accuracy of Breast Cancer. A comparative investigation was newlinecarried out in testing the dataset by adopting the parameters- with feature selection and without newlinefeature selection. Further, the analysis results in state that, the model with feature selection newlineobtained an enhanced accuracy of 0.74% than without feature selection. newlineFurther, the second contribution deals with developing the improvised feature selection newlinetechnique. The proposed scheme develops

Description

Keywords

Citation

item.page.endorsement

item.page.review

item.page.supplemented

item.page.referenced