US Patent No. 7,630,930 -- Method and system for portfolio optimization from ordering information
This invention involves the optimization of investment portfolios by finding the centroid of a cone in a very high-dimensional space formed by inequality relationships. A challenge in getting patents related to financial methods allowed is overcoming "non-statutory matter" rejections. In this case we tied the algorithm to the computer's processor and memory. The Patent Office's standards for what is allowable and what is considered non-statutory has been in flux for the last few decades. Recently, the Supreme Court ruled that “Merely requiring generic computer implementation [of a business method] fails to transform that abstract idea into a patent-eligible invention.”
A method of optimizing a portfolio includes selecting an investment universe with a finite number of assets, forming a belief matrix based on one or more homogeneous inequality relationships among the expected returns of assets in the universe, selecting those asset returns that are consistent with the belief matrix to form a consistent set of return vectors, selecting a set of allowable weight vectors for the assets in the universe, determining a centroid vector of the consistent set of return vectors with respect to a probability measure, and finding an optimal portfolio by finding a weight vector on a boundary of the set of allowable weight vectors that maximizes an inner product with the centroid vector.