Cancer Prediction Using Machine Learning

Breast cancer is a common form of cancer experienced in women across different age groups and accounts for 14% of cancers in Indian women. The high mortality rate is due to the lack of awareness of the disease and delay in screening and diagnosis. Early detection is the solution which implies that breast cancer can be treated, and patients have the chance to live a healthy life post-recovery. The proposed system aims at providing a diagnostic analysis of the cytological features derived from a biopsy-based image. The designed model attempts to predict cases of malignancy or benign on the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. The dataset used for analysis and implementation consists of attributes obtained from images of fine needle aspiration (FNA) tests on an affected breast tissue. The feature selection algorithm implemented was recursive feature elimination based on random forest approach. A comparative analysis was carried out against several other classification methods in an attempt to compare the accuracy of random forest approach against each of the other models. The neural network constructed using the features selected exhibited an accuracy of 96.51% while predicting malignant and benign cases in the test set. The main objective of this study was to devise an efficient model that a pathologist can effectively use in determining cases of malignancy based on the most critical cytological features. This ensures high accuracy, thereby enabling a patient to avail of the required medical attention at the earliest.