Multi-View Ensemble Classification For Clinically Actionable Genetic Mutations

This paper presents details of our solutions to the task IV of NIPS 2017 Competition Track that is called Classifying Clinically Actionable Genetic Mutations. It aims at classifying genetic mutations based on text evidence from clinical literature. A novel multi-view machine learning framework with ensemble classification models is proposed to solve this problem. During this Challenge, feature combinations deriving from three views including document view, entity text view, and entity name view to complement each other are comprehensively explored. Finally, an ensemble of 9 basic gradient boosting models win in the comparisons. Our approach scored 0.5506 and 0.6694 in Logarithmic Loss on a fixed split of stage-1 testing phase and 5-fold cross validation respectively, which is ranked as a top-3 team in NIPS 2017 Competition Track IV.