We discuss the use of machine learning algorithms to predict which breast cancer patients are likely to respond to (neoadjunctive) chemotherapy. A group of 96 patients from the Aberdeen Royal Infirmary had the size of their tumours assessed by Positron Emission Tomography at various stages of their chemotherapy treatment. The aim is to predict at an early stage which patients have low response to the therapy, for which alternative treatment plans should be followed. A variety of machine learning algorithms were used with this data set. Results indicate that machine learning methods outperform previous statistical approaches on the same data set.