Sanat Kumar1

1, Columbia University, New York, New York, United States

Machine Learning (ML) has been used for quite some time in analyzing large sets of data in a variety of disciplines in order to make predictions on new data. Although gas permeability data has been collected for polymer membranes for decades, only hundreds of data points are available in the literature out of the thousands of possibilities. In order to address the intractability of running thousands of different gas permeability studies, we use ML to build a model on existing data and predict which untested polymers are most likely to achieve superior gas separation performance.