This paper proposes the use of machine learning algorithims to model price/yield relationships in fixed income attribution. The advantage of the described technique is that it produces highly accurate and detailed analyses, but without the substantial data and pricing requirements needed by other approaches.