Breast Cancer Classification

Abstract: Acoustic For a limited number of disorders, early diagnosis of any illness may be curable to mankind's commitment. Before it becomes chronic, most persons fail to detect their illness. This adds to a global rise in mortality rates. Breast cancer is one of the cancers that can be treated until it progresses to all areas of the body as the condition is diagnosed at early stages. Breast cancer primarily affects women and it is also an important factor in raising the rate of female mortality. We are both mindful that the diagnosis of breast cancer is very time-consuming. On the other hand, the availability of technology used to diagnose early-stage cancer is very limited. Different algorithms for Machine Learning and Deep Learning have been used to distinguish benign and malignant tumors. UCI Wisconsin database containing 569 samples and 31 features is included in these papers. The paper focuses on numerous models that are applied to the dataset taken, such as K Nearest Neighbor, Vector Machine Support, Random Forest, Naive Bays, Logistic Regression, and Artificial Neural Network (ANN), etc. In terms of accuracy, cross-validation, sensitivity, and specificity gained, each of these algorithms was calculated and compared. From the experiments, we come to the solution that Random Forest give the best accuracy and the and the worst algorithm K-nearest Neighbors accuracy is 99. 20%. Deep learning algorithms ANN have been applied to improve prediction accuracy. The overall accuracy reached in the case of ANN 99.73%, respectively. There are two types of activation functions used for forecasting namely Relu and Sigmoid.

Introduction: Breast Cancer currently affects and destroys most women. This is because breast cancer is the world's second-most deadly cancer. 12% of all current forms of cancer and 25% of all women's cancers are breast cancer.

Tumors that can be cancerous are called the being out influence implementation of cells in an organ. Benign and malignant are two types of tumors that are available in the world. Non-cancerous cancers that do not spread and do not intimidate life are known as benign cancer. On the other hand, there is an enlarging and life-threatening malignant or cancerous tumor. Between the age of 40 and 55 women are affected and the leading cause of death because of Breast cancer.

Normal diagnosis of breast cancer followed by effective treatment for cancer may decrease the risk of unwillingness. It is recommended that tumor assessment tests be performed every 4-6 weeks. Benign and malignant distinction based on classified features has been very relevant on this basis [6].

It is proven that if anyone takes early diagnosis carefully, the rate of death is decreased. Depending on the experience, when recognizing a disease, medical professionals may make errors. We will make our diagnosis more precise (91.1 percent) by using technology such as data processing, deep learning, artificial learning, where the seasoned doctor found it (79.9 percent) correct. [5].

Nowadays, ANN has been one of the best deep learning models because of its data processing task likes classification and regression. We are assured that the diagnosis of breast cancer, ANN performs the best accuracy. Even though, the tectonic has many drawbacks. First, we have to tune the parameters of ANN. There are many parameters that we have to tuned namely learning speed, hidden layer, activation functions, etc. Second, due to the dynamic architecture and parameters update process, it needs more computational costs and it takes a long time for the training phase. Third, it can be stuck to a local minimum so that it is not possible to guarantee optimum efficiency.

For diagnosis and detection of breast cancer effects, there are several algorithms. The current paper compares the efficiency of five classifiers: SVM, NB, LR, RF, and k-NN, which are among the research community's most popular machine learning techniques and the top 10 machine learning techniques. Our goal is to assess the accuracy, sensitivity, specificity, and accuracy of the performance and quality of these algorithms.