Abstract: Cancer is one of the deadliest diseases and of all the cancers breast cancer is the prime reason of cancer death among women as compared to men, today 1 in 8 women are suffering from breast cancer according to the American Cancer Society (Information collected from cancer.org website https://bit.ly/3uc13B3). As of now breast cancer accounts for 41 % of cancer deaths in women and this number is likely to increase by 2030, according to the world health organization (Link: https://bit.ly/3ioT8eh) in 2020 there were 2.3 million women diagnosed with breast cancer and the death toll was 685 ,000. Some of the factors that are contributing to this disease are usage of alcohol, exposure to cigarette smoke, no or minimal breastfeeding, lack of physical activity, family history of breast cancer, obesity, gene change and exposure to radiation. Through this research paper we are trying to investigate breast cancer triggering factors and applying data mining models to work on early detection based on patients’ medical history and predicting where this disease will reoccur or not with accurate results. Our data mining model uses Recursive feature Elimination method with cross validation, Feature Selection and stacked with Artificial Neural Network to detect breast cancer. Furthermore, we have also compared the Artificial neural network classifier with other Machine learning algorithms to find the accuracy. To add some statistics into our model we have used concepts like Specificity, Sensitivity, ROC-AOC Score, Kappa-Cohen score, Wilcoxon Signed Rank test and other statistical parameters to check and compare the Artificial Neural Network model with other Machine Learning Classifier Algorithms. In our model we have represented other Machine learning classifier algorithms as Logistic Regression, Decision tree, Random Forest classifier, K-Nearest Neighbor, Support Vector Machine, Gradient Boosting and Naive-Bayes. In this Paper we have Proposed a stacked model which can outperform other Machine Learning Classifier Algorithm by calculating various statistical parameters and by conducting non-parametric test to prove our hypothesis.
In this research Paper the authors observed 98 percent accuracy by using Artificial Neural Network Based Approach along with Recursive Elimination Feature Selection combined model with Hyperparameter Tuning so that we observed sensitivity being 100 percent and specificity 99 percent also the ROC-AUC Score is 100 percent and the kappa score is 99 percent.
Keywords: Machine Learning, Deep Learning, Breast Cancer, Feature Selection
|
DOI:
10.17148/IJARCCE.2022.11335
