A neural network based breast cancer prognosis model with PCA processed features

Accurate identification of the diagnosed cases is extremely important for a reliable prognosis of breast cancer. Data analytics and learning based methods can provide an effective framework for prognostic studies by accurately classifying data instances into relevant classes based on the tumor severity. Accordingly, a multivariate statistical approach has been coupled with an artificial intelligence based learning technique to implement a prediction model. Principal components analysis pre-processes the data and extracts features in the most relevant form for training an artificial neural network that learns the patterns in the data for classification of new instances. The diagnostic data of the original Wisconsin breast cancer database accessed from the UCI machine learning repository has been used in the study. The proposed hybrid model shows promising results when compared with other classification algorithms used most commonly in the literature and can provide a future scope for creation of more sophisticated machine learning based cancer prognostic models.