Predictive Analytics for Breast Cancer Survivability: A Comparison of Five Predictive Models

Predicting the statistics of survival from the innumerable reports of breast cancer is a difficult job to perform, one that every researcher tackles in the field of medicals science. Since the birth stages of the related research, many new techniques have advanced. This point is evident from the groundbreaking biomedical technologies and discovery of better self-explanatory predictive factors. Due to supply of cheaper computer hardware and easy availability of software technologies, large amount of qualitative data is being stored automatically. The developed ways of analyzing huge data quantities have bestowed upon this field efficient and effective means of processing. Five renowned data mining algorithms (Naive Bayes classification, decision trees, IBK, support vector machine (using SMO) and OneR,) were used for our prediction models applied to a substantiate dataset (more than 683cases). The10-fold cross-validation method was also taken into due consideration to determine the prediction accuracy of the five prognostic designs in comparing their credibility. Through a sensitivity analysis of support vector machines, we were able to gauge the prioritized importance of predictive factors in this work.