Study on Constructing Breast Cancer Prediction Model with Different Algorithms

Nowadays breast cancer had become one of the most serious cancers for women in the world. Therefore, constructing a risk prediction model for breast cancer is of great importance. This paper focuses on the goal to construct a best breast cancer prediction model based on SVM, KNN, and Naïve Bayes algorithms, which is constructed based on dataset from Wisconsin Breast Tumor dataset that has 569 samples with 30 calculated traits derived from digital photographs of the cell nuclei. For methods used in the paper, data-preprocessing, EDA, PCA are used to process the dataset. After that, prediction models are constructed and optimized based on three algorithms. Results show that SVM are the best option for the dataset in this study, according to the experiment, however KNN is better suited for datasets with more samples or samples that will be separated such that the class fields are heavily crossed or overlapped. For samples with fewer features, the Nave Bayes method is more suited.