There are several reasons why predictive modeling for cancer has become an important use case. For one thing, cancer is the second leading cause of death among medical diseases, just behind heart attacks. It's insidious onset and course makes cancer diagnosis just that bit more surprising and devastating. No one can dispute the importance of fighting cancer with every tool in our arsenal, and that includes machine learning methods.

Second, within cancer machine learning, there are a variety of use cases that are well-suited to being solved by machine learning. For example, given a healthy patient, how likely is that patient to develop a particular type of cancer? Given a patient that has just been diagnosed with cancer, can we inexpensively predict whether the cancer is benign or malignant? How long can the patient be expected to survive? Will they likely be alive in 5 years? 10 years? To which, chemotherapy/radiotherapy regimen is the patient most likely to respond? What is the chance of cancer recurring once it is successfully treated? Questions like these benefit from mathematical answers that may be beyond the capabilities of a single doctor's reasoning or even that of a panel of doctors.